Language, Sense and Nonsense: A Critical Investigation into Modern Theories of Language

"�--�-----"-

-------�- --- -------------- ----------

Language, Sense and Nonsense Critical Investigation into Modern Theories of Language A

�

G. P. Baker & P. M. S. Hacker Fellows of St John's College . Oxford

Basil Blackwell

© G. P. Baker and P. M. S. Hacker 1984

First published 1984 Basil Blackwell Publisher Limited 108 Cowley Road, Oxford OX4 1]F, England All rights reserved. Except for the quotation of short passages for the purposes of criticism and review, no part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher.

British Library Cataloguing in Publication Data Baker, G. P. Language, sense and nonsense. 2. Languages - Philosophy 1. Linguistics I. Title II. Hacker, P. M. S. 401 P121 ISBN 0-63 1-135 19-7

Typeset by Oxford Verbatim Limited Printed in Great Britain by T.]. Press Ltd, Padstow

For Alan and Jonathan

La direction de notre esprit est plus importante que son progres

Preface

The emperor walked in the procession under his crimson canopy. And all the people of the town, who had lined the streets or were looking down from the windows, said that the emperor's clothes were beautiful. 'What a magnificent robe! And the train! How well the emperors clothes suit him!' None of them were willing to admit that they hadn't seen a thing; for if anyone did, then he was either stupid or unfit for the job he held. Never before had the emperors clothes been such a success. 'But he doesn't have anything on!' cried a little child. Hans Christian Andersen

Between 1 978 and 1 9 8 1 we busied ourselves investigating the origins of modern logical theory and the inspiration for much contemporary philosophical logic in the works of Gottlob Frege. The results of our historical investigations were startling, for the picture of Frege's thought that emerged from our labours diverged in many fundamental respects from what might be called the standard average conception of his philosophy. The roots of modern logic did not lie in semantic investiga­ tions into the structure of natural languages, but in his application of sophisticated mathematical techniques to the traditional subject-matter of logic, viz. the nature of concepts, judgments and valid inference. What Frege had done was to invent a new and powerful method of representa­ tion for presenting the forms of judgment and valid inference which are recognized and expressed in natural languages and in mathematics. What his successors did was to project this novel form of representation back on to natural languages, the mind or the world in the guise of theories about the 'depth structure' of any possible language, the 'logical form' of thoughts, or the logical structure of the states of affairs described. The fruits of our labours we presented in Frege: Logical Excavations, in which we analysed the function-theoretic mainspring of his logical theory and subjected its philosophical foundations to thorough criticism. This exercise in philosophical archaeology left much important unfinished business. First, establishing the formal and ultimately mathematical inspi­ ration of his reflections had the effect of divorcing much of the criticism of his main ideas from a critical analysis of the major conceptions informing modern philosophy of language (many of which are wrongly attributed to Frege by most authors). Secondly, making out a case for a predominantly negative verdict on the coherence, intelligibility, and interest of his work for modern philosophical and logical investigations might falsely suggest that we were picking a quarrel with a past genius for not matching present-day standards of sophistication. In fact we do consider that the

Preface

Vlll

guidelines of modern philosophical logic or philosophy of language are worth careful analysis in their own right, and we think ourselves to have batteries of arguments available to demonstrate that modern wisdom is in many fundamental respects not at all superior to Frege's (different) con­ ception. We had a double motive to investigate modern' theories of language. To that task the present book is dedicated. Contemporary philosophers, competing in the market-place of ideas, have produced a wide range of different theories about the nature of a language, of mastery of a language, of thought and understanding. Modern theoretical linguists present as diverse a spectacle of wares. Passionate controversies rage between different brands of transformational genera­ tive grammarians, truth-conditional theorists, speech-act theorists, and many other doctrinaires. To confront each seriatim would be both lengthy and unprofitable. Amidst the roar and hubbub of conflicting voices we have tried to discern not the differences between the competing theories, but the agreed presuppositions, the common Idols of the Market-place and the accompanying Idols of the Theatre' set up in the colourful adjoining booths. That there is wide agreement (not, of course, universal, or uniform) can hardly be doubted. Most theorists view the fact that a speaker of a language can understand sentences he has never heard before as a deep observation calling out for explanation. Many theorists concur in the conception of a language as a calculus of rules for the use of symbols. These hitherto unknown rules determine the grammaticality of combina­ tions of words, as well as the senses they convey. Speaking and underI Cf. F. Bacon: 'There are also idols formed by the intercourse and association of men with each other, which I call Idols ofthe Market-place, on account of the commerce and consort of men there. For it is by discourse that men associate; . . . And therefore the ill and unfit choice of words wonderfully obstructs the understanding. Nor do the definitions or explanations wherewith in some things learned men are wont to guard and defend themselves, by any means set the matter right. But words plainly force and overrule the understanding, and throw all into confusion, and lead men awry into numberless empty controversies and idle fancies. ' 'Lastly, there are idols which have immigrated into men's minds from the various dogmas of philosophies . . . These I call Idols of the Theatre; because in my judgment all the received systems are but so many stage plays, representing worlds of their own creation after an unreal and scenic fashion. Nor is it only of the systems now in vogue, or only of the ancient sects and philosophies, that I speak: for many more plays of the same kind may yet be composed and in like artificial manner set forth; seeing that errors the most widely different have nevertheless causes for the most part alike. Neither again do I mean this only of entire systems, but also of many principles and axioms in science, which by tradition, credulity and negligence have come to be re ved.' (Novum Organum, xliii-xliv)

g{:

Preface

IX

standing a language is commonly conceived as a matter of operating this complex calculus of precise rules, even though speakers have no conscious knowledge of them. Only thus can one explain and predict the limits of grammaticality and the bounds of sense, as well as render intelligible the mysterious processes of understanding. It is precisely because a language is a calculus of rules that it is possible to understand new sentences. The task of the theorist of language is to discover the forms of these rules, and thus to lay bare for the first time the hidden structures of languages. The task of the logical theorist is to demonstrate that the validity of the inferences we correctly take to be valid is explained and guaranteed by the underlying rules of a language. This modern conception of a language is the work, not of one pair of hands, but of many. It lies at the confluence of numerous different streams of thought. The invention of modern mathematical logic at the hands of Frege and his successors gave philosophers a particular vision of the underlying logical structure of language, thought and reasoning. The development of post-Saussurean synchronic linguistics has tended to move along parallel tracks, thus reinforcing that vision. The development of computers, the operation and programming of which rests on logico­ mathematical principles, gave further impetus to the idea that the human mind works on the same principles as its recent favoured brain-child. Advances in neurophysiology have taken inspiration from the much over­ worked analogy between the functioning of the brain and the operations of a computer. And these in turn served to reinforce a variety of trends in modern philosophy of mind as well as empirical psychology towards mind-brain identity theories or computational functionalism. If some version of the predicate calculus is not, as Russell and the young Wittgen­ stein thought it was, a mirror of the logical structure of reality, at least it seems that it (or at any rate the depth-grammar of a natural language) must be a mirror of the logical structure of the mind and of the functional structure of the cognitive capacities of the brain. If the favoured pictures of language are, as we argue in this book, misconceived, the misconceptions are deep and numerous. They are held firm not by one fallacy but by many mutually reinforcing fallacies. And given the multiplicity of their sources, they are not refutable by frontal assault. But where the curtain walls and bastions of a fortress are impreg­ nable, the foundations may yield to undermining. We have chosen four related points from which to tunnel, four fundamental themes command­ ing widespread consensus. These are, first, the doctrine of the separation of the sense of a sentence from its force; secondly, the conception of the

x

Preface

truth-conditions of a sentence as the key to a comprehensive theory of meaning; thirdly, the notion of a hidden, 'tacitly known', system of linguistic rules underlying our thought and speech; and finally, the alleged mystery of our capacity to understand sentences we have never heard before. If our strictures on these themes are right, then most of what goes by the name of 'theories of meaning' or 'scientific study of language' needs not remedial readjustment, but wholesale abandonment. The issues we examine are important, not only for philosophy and the philosophical understanding of the nature of thought and language, but also for empirical linguistics and psychology. The misconceptions we identify ramify widely, contributing greatly to the barren mythology of late twentieth century culture. Hence this book is written with more polemical passion than is common in the typical reserved and detached forms of academic philosophy. For this we make no apology. It would be foolish to hope, let alone to expect, that a great structure, as attuned to the spirit of the age as modern theories of meaning in philosophy and theoretical linguistics, with their richness, diversity and complexity, will crumble as a consequence of our efforts. The most that we hope for is not the immediate collapse of the ramparts, but rather that some of those who man them, in particular prospective recruits who have not yet taken the Queen's shilling, will abandon them in the realization that for all the apparent power of the structure, the foundations rest on sand. But perhaps even this is too optimistic, in which case we can only endorse the observation that 'there is nothing, so plain boring as the constant repetition of assertions that are not true, and sometimes not even faintly sensible; if we can reduce this a bit, it will be all to the good'. 2 We are happy to record our gratitude to Dr John Dupre, Bede Rundle, Stuart Shanker and Dr Talbot Taylor, who read and commented on several draft chapters of this book. We benefited from numerous conversa­ tions with them. Our greatest debt is to Professor Roy Harris, who guided us through many of the thickets of modern linguistic theory, patiently answered our endless questions, and most helpfully criticized drafts of many chapters. His kindness, wit and encouragement sustained us in hours of gloom. Our college, St John'S, generously reduced our teaching load while we were composing this book and gave us liberally of its many facilities. G. P. B. & P. M. S. H.

St John's College, Oxford 1983

2 J. L. Austin, Sense and Sensibilia (Clarendon Press, Oxford, 1962), p. 5.

Contents

Prolegomenon: the New Philosopher's Stone CHAPTER 1

1 2

3 4

5

The post-Cartesian tradition The conception of language within the post-Cartesian tradition The logical tradition The invention of formal calculi The watershed

CHAPTER 2

1 2

3

4

3 4

5

Sense and Force: the Evolution of the Species

The background environment Phylogenesis Lateral radiation Suspicions of genetic defects

CHAPTER 3

1 2

Historical Bearings

Sense and Force: the Pathology of a Species

Descriptive content Overlapping paraphrases Shortage of logical space An indirect approach via indirect speech Diagnosis

1 14 14 17 25 30 39

47 47 54 69 76

80 80 87 94 106 1 13

Truth-Conditions: Origins and Evolution

121

The present position Logical semantics 3 Truth-conditions domesticated: applications to natural languages 4 Forging links between language and the world 5 Cutting language free from actuality 6 Diaspora 7 The evolution of a concept

121 126

CHAPTER 4

1 2

132 140 147 153 162

CHAPTER 5

1 2 3 4 5

1

The Mythology of Rules

Methodological confusions The psychological mythology of rules The mechanical mythology of rules The Platonic mythology of rules The mythology of scientific method

CHAPTER 9

1 2 3 4 5 6 7

Rules: Preliminary Clarifications

Clouds of confusion Rules and their formulations Normative phenomena The existence of rules and the determination of their consequences

CHAPTER 8

2 3 4 5

Truth-Conditions: Ramifying Defects

In pursuit of a mirage Synonymy Circumstance-dependence of meaning Sentence-meaning Explanation Understanding In pursuit of the phantoms of hope

C HAPTER 7

1 2 3 4

Truth-Conditions: Flaws in the Foundations

Targets Semantic analyses of logical constants Truth and falsity Truth-conditions Context-dependent sentences

CHAPTER 6

1 2 3 4 5 6 7

Contents

Contents

xu

The Generative Theory of Understanding

Plus �a change . . . The plants that grew The elements Combinations of elements: phrases and sentences Understanding and tacit knowledge Understanding new sentences Residual business: the minimalist enterprise

168

CHAPTER 10

168 170 180 190 197

1 2 3 4 5

206

Index

206 207 218 228 232 238 241 243 243 250 256 262 267 267 286 294 300 307 316 316 321 327 334 339 345 356

Turning Full Circle

Kernels of truth? Discredited methodology A mythology of symbolism Forms of representation Philosophy of language

Xlll

369 369 371 373 379 386 390

Prolegomenon: the New Philosopher's Stone

Since the dawn of their subject twenty-five centuries ago, philosophers have sought to clarify fundamental notions such as appearance and reality, substance and property, truth and falsehood, justice and virtue, notions which play a crucial role in our understanding of the world in which we live, of ourselves as fleeting inhabitants of it, and of the goals and values which we pursue and which give meaning to our lives. Throughout its long history, they have constantly deployed arguments of a broadly linguistic nature in pursuit of their aims. Socrates, in his endeavour to understand the nature of the virtues, sought for clear definitions which would capture the essence of courage or friendship, of justice or wisdom. And proposed definitions are examined, sometimes rejected, sometimes modified, in the light of what we would, in various circumstances, find it appropriate to say. Aristotle's subtle discussions of the voluntary and involuntary, of choice and deliberation, of responsibility and culpability, abound in argu­ ments showing that one kind of thing cannot in general be identified with another kind of thing because what can legitimately be said of one cannot be said of the other, or because the name of the one cannot be exchanged for the name of the other in an utterance without a change in what is said. Such arguments can fairly be characterized as being, in a loose sense, linguistic. Nor is this all. From the beginnings of philosophy, writers have sought to clarify the nature of thinking and reasoning, to codify the canons of valid argument and to demonstrate their ultimate grounds in the nature of the thinking mind or in the nature of the realities thought about. Here too philosophers have always viewed language as a guide to the nature of the thoughts and arguments expressed. It was, to be sure, commonly viewed as a wayward guide, pointing to a reality which underlies it, and not always accurately reflecting what it thus represents. But by cautiously and criti-

Prolegomenon

Prolegomenon

cally following its lead, the logician can learn much about concepts and judgments, valid arguments and the nature of inference. To be thus concerned with language, to deploy such linguistic argu­ ments and to see the grammar of language and the structure of sentences as valuable clues to the nature of thought is one thing. But to declare that language actually provides part of, perhaps indeed the main part of, the subject-matter of philosophy, is another thing altogether. For past genera­ tions the investigation of a fairly narrow range of features of language, of different kinds of words and their distinctive roles, of the grammatically licit forms of combinations of words into different kinds of sentences, was an important preliminary undertaking in preparation for larger philo­ sophical enterprises. Linguistic and grammatical data provided tools in the philosophical workshop, tools which could be put to use, by a skilled craftsman, in pursuit of solutions to fundamental philosophical questions. But it was only relatively recently that some philosophers began to con­ ceive of the philosophical study of language, not merely as a tool, but as a task, indeed as the primary task of philosophy. This shift in philosophers' conception of their subject-matter was pro­ duced to a large extent through the examination of the apparent philosophical implications of the discovery, in the nineteenth and early twentieth centuries, of rich and sophisticated logical calculi, These formal inventions were the fruits of the labours of philosopher-mathematicians such as Boole, Frege, and Russell and Whitehead. The actual transforma­ tion in the conception of philosophy can be dated, with fair accuracy, to Wittgenstein's Tractatus Logico-Philosophicus ( 1 921). In it he argued that the hidden structure of any possible language is akin to that of a (correct) formal, function-theoretic, logical calculus. He gave a wholly original and in many ways deeply satisfying conventionalist explanation of the nature of logical truth, dissolving logical necessity into conventions for the employment of symbols. And he argued that the sole task of philosophy is the clarification, by logical analysis, of sentences of natural language, and the eradication of pseudo-propositions (in particular, metaphysical ones) which violate the logical syntax of language. 'All philosophy', he declared, 'is a critique of language'. I These ideas caught on, particularly among the logical positivists in the 1 920s and 1 930s. Schlick, the leader of the Vienna Circle, declared:

I am convinced that we now find ourselves at an altogether decisive turning point in philosophy . . . Leibniz dimly saw the beginning. Bertrand Russell and Gottlob Frege have opened up important stretches in the last decades, but Ludwig Wittgenstein (in his Tractatus Logico-Philosophicus, . . ) is the first to have pushed forward to the decisive turning point.2

2

I L. Wittgenstein, Tractatus Logico-Philosophicus, tr. D. F. Pears and B. F. McGuinness (Routledge & Kegan Paul, London, 1961), 4.003 1 .

3

.

This sentiment was shared by members of the Circle,3 and their conception of philosophy as the analysis of the language of science (in a broad sense) and the extirpation of metaphysics rested on what they conceived to be the central lessons of the Tractatus. 'All philosophical problems', Carnap boldly announced, 'are questions of the syntax of the language of science. '4 Schlick proclaimed that 'Investigations concerning the human "capacity for knowledge" . . . are replaced by considerations regarding the nature of expression, of representation, i.e. concerning every possible "language" in the most general sense of the term.'5 Ayer declared that the function of philosophy 'is to clarify the propositions of science, by exhibiting their logical relationships, and by defining the symbols which occur in them'6 . . . 'A complete philosophical elucidation of any language would consist, first, in enumerating the types of sentence that were significant in that language, and then displaying the relations of equivalence that held between sentences of various types. '7 This conception went into eclipse in the 1950s and 1 960s when, paradoxically, a movement popularly (and misleadingly) known as 'lin­ guistic philosophy' was in its heyday. That 'movement' spurned common goals or ideology in the grand manner of the Vienna Circle. If it can be generally characterized at all, it is primarily by its employment of linguistic methods, rather than by its avowal that the investigation of natural language was the subject-matter of philosophy, or that a subject called 'philosophy of language' was the foundation of the whole of philosophy. Its leading representative, J. L. Austin, explicitly declared that 'ordinary language is to be our guide', that proceeding 'from "ordinary language", 2 M . Schlick, 'The Turning Point in Philosophy' in Logical Positivism, ed. A.J. Ayer (Free Press, Illinois, 1 959), p. 54. 3 Cf. their manifesto 'The Scientific Conception of the World: The Vienna Circle'. 4 R. Carnap, 'On the Character of Philosophical Problems' (1934), in The Linguistic Turn, ed. R. Rorty (University of Chicago Press, Chicago, 1967), p. 61. 5 M. Schlick, 'The Turning Point in Philosophy', p. 55. • A. J. Ayer, Language, Truth, and Logic, 2nd edn (Pelican, Harmondswotth, 1971), p . 42. 7 Ibid., p. 83.

4

Prolegomenon

that is by examining what we should say when . . . [is] . . . one philosophical method'.8 Practitioners of this method were not given to large-scale theory-construction, nor to grandiloquent pronouncements about systematic theories of meaning. They did not conceive of philosophy as continuous with the empirical sciences, nor did they conceive of its task as the analysis, reconstruction or regimentation of the language of science. They were united only by a common belief that many philosophical problems could be resolved by linguistic methods, that many philosophi­ cal puzzlements arose out of misunderstandings concerning our use of words, and by a piecemeal rather than holistic approach to philosophical issues. Ryle's Concept of Mind was not conceived as an investigation into the logical syntax of the language of psychological science (as Carnap might have conceived of such a project) but as a study of 'the logical geography' of ordinary mental concepts. Austin described his essay 'A Plea for Excuses' as a branch of the philosophical study of conduct, 9 and Hart declared that his Concept of Law aimed 'to further our understanding of law, coercion and morality'. Though it employed analytical linguistic methods, it was, he insisted, also an essay in 'descriptive sociology'. 10 He quoted with approval Austin's remark that we may use 'a sharpened awareness of words to sharpen our perception of, though not as the final arbiter of, the phenomena'. II Towards the end of the 1 960s, the waves of 'ordinary language philosophy' gradually subsided. And although the central contentions of the Tractatus were explicitly disowned by the older Wittgenstein, the seeds sown by his younger self and fostered by the logical positivists, sprouted afresh. In the late 1 960s and throughout the 1 970s a second generation of philosophers flooded the market-place with ideas explicitly or tacitly harvested from the fields of the Tractatus and the logical positivists. 1 2 The first and foremost task of philosophy, it was proclaimed, is to construct a theory of meaning for a natural language, to elicit the underlying prin­ ciples of construction of any language in virtue of which we can construct and understand the infinite array of meaningful sentences with which we can express our thoughts. M. A. E. Dummett, in Oxford, maintained that 8 J. L. Austin, 'A Plea for Excuses', in Philosophical Papers, 1st edn, ed. J. O. Urmson and G. J. Warnock (Clarendon Press, Oxford, 1961), p. 129. 9 Ibid., p. 128. 10 H. L. A. Hart, The Concept ofLaw (Clarendon Press, Oxford, 1961), p. vii. II Austin, Philosophical Papers, p. 130. Il Although by a curious misreading of history, they tended to conceive of Frege as their honorary ancestor (see pp. 32-9).

Prolegomenon

5

'philosophy has, as its first if not its only task, the analysis of meanings', 13 that 'the philosophy of language is the foundation of all other philosophy', 14 and that 'the most urgent task that philosophers are now called upon to carry out is to devise . . . a "systematic theory of meaning" '.15 Indeed, 'If we had an agreed theory of meaning, then that theory could be appealed to in order to find a resolution of [the] problems [of philosophy in other localized areas, e.g. metaphysics or philosophy of mathematics].' 16 A philosophical theory of meaning is here advertised as the true Philosopher's Stone. A similar vision captured the imagination of some American philos­ ophers. D. Davidson declared: I dream of a theory that makes the transition from the ordinary idiom to canonical notation purely mechanical, and a canonical notation rich enough to capture, in its dull and explicit way, every difference and connection legitimately considered the business of a theory of meaning. The point of a canonical notation so conceived is not to improve on something left vague and defective in natural language, but to help elicit in a perspicuous and general form the understanding of logical grammar we all have that constitutes (part of) our grasp of our native tongue. 17

And he too thought that remarkable consequences would flow from the realization of this dream. For example, a correct semantic analysis of action- and event-sentences will, he suggested, constitute a proof that events actually exist. 1 8 Ontological truths can thus be proved from pre­ mises provided by a theory of meaning. The renaissance of interest in global questions about the nature of language and the renewal of effort to construct comprehensive theories of meaning has been stimulated by a fascination with the fact that language­ users are able to understand a potentially infinite array of sentences never heard before (a thought explicit in the Tractatus). Since what a language­ user learns is surely finite, these resources must somehow suffice for the 13 M. A. E. Dummert, Frege: Philosophy of Language (Duckworth, London, 1973), p. 669. 14 M. A. E. Dummett, 'Can Analytical Philosophy be Systematic and Ought it tp be?', in Truth and Other Enigmas (Duckworth, London, 1 978), p. 442. 1 5 Ibid., p. 454. 1 6 Dummett, Frege: Philosophy of Language, p. 672, d. p. 676. 1 1 D. Davidson, 'The Logical Form of Action Sentences', in The Logic of Decision and Action, ed. N. Rescher (Universiry of Pittsburgh Press, Pittsburgh, 1967), p. 1 15. 18 A bizarre view, given that events neither exist nor fail to exist, but rather occur, happen or take place. It is difficult to see why one would want a proof that things happen.

6

Prolegomenon

generation of, and possibility of understanding, any of the infinite sen­ tences of this language. This alone seems to prove that a speaker has an implicit grasp of a variety of constructional principles of language, knowledge of which will explain this capacity. And an explicit statement of these principles, which displays how to derive the meanings of sentences from the meanings of constituent words and their principles of composi­ tion into sentences, is precisely what a complete theory of meaning for a language is. Philosophers see in this reasoning a 'transcendental deduc­ tion' of the existence of a theory of meaning. Despit� many differences in the methods and products of theory con­ struction, it is possible to give a rough sketch of typical theories of meaning (of the simple type). More elaborate accounts will be given later. The theorist distinguishes sharply between three enterprises: syntax (the sci­ ence of the principles for constructing grammatical sentences), semantics (the study of the truth-conditions of sentences) and pragmatics (the in­ vestigation of all those features of sentences related to their use by speakers in particular contexts) . Employing the familiar distinction between type­ and token-sentences, he focuses upon the type-sentence as the primary bearer of meanings, relying on pragmatics to bridge the gap between the meaning of the type-sentence and the description of what a speaker has done in uttering one of its tokens on a given occasion. The type-sentence is conceived as having priority over words in the order of semantic analysis, its meaning being thought of as given by its truth-conditions, viz. a specification of the conditions under which instances of it can be used to propound a truth. Its constituent words are conceived as having a meaning which consists in their contribution to determining the truth-conditions of sentences in which they may occur. So the meaning of a (type-)·sentence (its truth-conditions) is viewed as a function of the meanings of its constituent expressions and of its structure, the manner in which the words are concatenated to form a sentence. Compound sentences are then conceived as built up (systematically) by operations on simple sentences. The picture underlying this theory-construction is a two-tiered explanation of language-use. An adequate syntax for a language should, when supple­ mented by a lexicon specifying the meanings of its words (and idiomatic phrases), assign a definite meaning to every well-formed sentence. This semantic theory, when supplemented by a specification of the relevant context of utterance of a sentence-token, should determine exactly what a speaker has done in uttering this token sentence (whether he has made an assertion, issued an order, etc., and also what he has asserted, ordered, etc.) . Since a theory of meaning aims at perfect generality, and since the

Prolegomenon

7

notion of a truth-condition is obviously tailored for declarative sentences typically used to make assertions, it is commonly thought that a sine qua non of such a theory is to distinguish between force (e.g. assertoric, interrogative, imperative force) and meaning (or sense). Every significant sentence is held to be analysable into a force-indicating device, e.g. 'It is the case . . . ' 'Is it the case . . . ?' 'Make it the case . . . ', and a truth-value bearing component, e.g. 'that so-and-so'. Sentences with different forces may yet have a common component, a description of a state of affairs, whose sense is given by its truth-conditions. This state of affairs may be asserted to obtain or ordered to be brought about, or one may query whether it obtains. In general, the underlying logical structures of the descriptive component of type-sentences are held to be given, mirabile dictu, by (some version) of the function-theoretic forms of the predicate calculus invented early in this century. This new philosophical vision has not occurred in a cultural vacuum. A common climate of thought has fostered parallel developments in linguistics. The resultant theories in the two subjects have reinforced each other. Modern 'structuralist' linguistics originated with the work of Saussure at the turn of the century. In sharp reaction to nineteenth century historical linguistics, he distinguished between the historical or diachronic study of language and synchronic study. The latter, conceived as an instantaneous cross-cut of the former, 1 9 is the domain of the science of language. Language (La langue) he viewed as an abstract structure of relations which govern the overt manifestations of speech (la parole) in the activities of language-users. It is the former which concerns the linguistic theorist. Although la langue was conceived, in Durkheimian fashion, as a supra­ individual social 'construct', it also had to be thought of, in some sense, as 'existing virtually inside everyone's head'. It was held to consist of a finite array of uniquely identifiable signs, governed by determinate rules of combination, and each sign (signi{iant) was correlated by a (theoretical) 'lexicon' with a meaning (signi{ie). Saussure thought of the relata as acoustic images and concepts. His successors modified this in various ways, but retained the 'biplanar'2o model of correlation, viz. of form and 'interpretation'. This conception of language involved a bold abstraction at more than one level. A language was not conceived as primarily a social practice, the existence of which is eminently visible (if not easily survey­ able) in the temporal stream of the activities of language-users, employing '9 For extensive criticisms of this very idea, see R. Harris, The Language Myth (Duck­ worth, London, 1981) and The Language Makers (Duckworth, London, 1980). 20 Cf. Harris, The Language Myth, p. 1 1 .

8

Prolegomenon

signs for a huge variety of human purposes in the context of ramifying auxiliary behaviour against the background of our familiar complex ma­ terial and social realities. These mundane features of historical develop­ ment, temporality, social practice, individual acts of speech, conventional and natural purposes, behaviour, context, etc. were, of course, recognized. But they were viewed as consequent upon the true nature of a language as an abstract, self-contained sign-system with a psychological reality of its own. The resultant conception of the science of language removed it from the sphere of socio-historic studies, allocating it instead to a branch of cognitive psychology. But it was successfully insulated from empirical psychology inasmuch as its subject-matter was held to be a wholly self­ contained abstract structure amenable to study only by linguistic tech­ niques. For its only observable manifestation amenable to experimental psychological methods is overt speech (la parole) which itself must be viewed as a consequence of the internalized structures of la langue. The new linguistics planted its flag on virgin territory, confident in the strength of its defences against encroachment from adjacent academic disciplines. After a brief behaviourist interlude, led by Bloomfield in America in the 1 93 0s and 1 940s, Saussurean structuralism (like the programme of the Tractatus) received a new lease of life (on a much larger scale) in the 1 960s and later. It was in effect revitalized, modified and extended by the development of transformational grammars at the hands of Chomsky and his followers. Like Saussure, they conceived of linguistics as essentially synchronic. Parallel to Saussure's distinction between la langue and la parole, they differentiated between a speaker's competence (his tacit knowledge of the system of language which the linguist studies) and his performance (the overt (and often faulty!) manifestations of his com­ petence). The grammatical rules they sought to elicit, as well as the abstract lexicon assigning 'interpretations' to signs, were conceived as 'internalized' in the mind or encoded in the brain, although (conveniently) inaccessible to introspection or to current neurological investigation. Their primary innovations lay in the development of the distinction be­ tween surface and deep structures of sentences, in the construction of syntax on the basis of transformations, and in the priority assigned to the sentence over sub-sentential expressions for purposes of grammatical theory. All of these features of modern linguistics have close parallels in the reflections of modern philosophers. The appearance of convergence between the theories of linguists and philosophers is striking. At first sight nothing seems more improbable than that ideas inspired by the Tractatus should have extensive parallels among

Prolegomenon

9

those inspired by Saussure. Surely the most obvious explanation is that there is an important niche in the ecology of the intellect which earlier evolution of ideas has left empty and which is now in the process of being occupied by independently evolved organisms. This seductive picture can be somewhat weakened by noting certain superficialities in the convergence. Linguists, lacking any direct check on the deep structures of sentences, are at liberty to borrow any advanced mathematical notions for characteriz­ ing deep structures, and they are strongly inclined to draw on the very same general ideas (sets, functions, quantifiers, variables) that are built into modern formal logic. Similarly, in considering meaning as a theoreti­ cal notion, they feel free to identify sentence-meanings with truth­ conditions and to treat this as the foundation of the science of semantics. Contrariwise, philosophers have drawn a blank cheque on the notion of logical form in stressing that grammatical form need not coincide with logical form, and therefore they are willing to cash this cheque in terms of the concept of depth-structure provided this manoeuvre promises to bestow some scientific cachet on their reflections. These points of converg­ ence nicely manifest intellectual opportunism. It is no surprise at all that philosophers and even some linguists are inclined to view the deep structure of language to be an elaboration and enrichment of the logical forms embodied in the predicate calculus. There is little to wonder at in the declaration 'It is rather generally supposed that we shall arrive at a satisfactory syntactic analysis of natural language only by exhibiting its sentences as having an underlying (or deep) structure analogous to that of sentences of Frege's formalized language [viz. the predicate calculus].'21 What is genuinely surprising is that linguists and philosophers are now so widely in agreement about the aims of their theory-construction and also about the methods appropriate for achieving these aims. Two prominent linguists declared that: Empirical linguistics takes the most general problem of the study of language to be that of accounting for the fluent speaker's ability to produce freely and understand readily all utterances of his language, including wholly novel ones. To explicate this ability, linguists construct a system of description which seeks to capture the regularities of the language used by speakers to produce and interpret sentences.22

This precisely mirrors the philosopher's inspiration: 21 Dummett, 'Frege's Distinction between Sense and Reference', in Truth and Other Enigmas, p. 1 1 8. 22 J. Fodor and J. J. Katz, 'What's wrong with the Philosophy of Language?', in Philosophy and Linguistics, ed. C. Lyas (Macmillan, London, 1971), p. 281.

10

Prolegomenon

Prolegomenon

a satisfactory theory of meaning must give an account of how the meanings of sentences depend upon the meanings of words. Unless such an account can be supplied for a particular language, . ..there would be no explaining the fact that we can learn the language: no explaining the fact that, on mastering a finite vocabulary and a finitely stated set of rules, we are prepared to produce and to understand any of a potential infinitude of sentences. 23

the great problems of philosophy, if grammar holds the key to the structure of the human mind, then indeed this wonderful insight and advance must be hailed with fanfares. And philosophers, together with theoretical linguists, must bend their wills to a united effort to grasp this treasure. Then they may go on to explain the deep mysteries of our ability to understand new sentences, to discover what really exists (e.g. whether events are essential for our 'ontology'), to reveal what is innately known to the human mind, to uncover the true logical form of our thoughts and the essential nature of our understanding. But the Last Trumpet has been blown with tiresome regularity in the history of philosophy, and false prophets have been legion. If the promises held out by the possibility of constructing a theory of meaning are false promises, and if the very idea of such a theory of meaning as is currently envisaged is incoherent, then this too must be proclaimed, the incoherences made clear and the hopes dashed. For then, far from being at last upon the true path of a science, theorists are merely pursuing yet another monstrous chimera. Lacking a true Philosopher's Stone, they will be in dire need of a Philosopher's Egg, a panacea against diseases of the intellect. In this book we shall subject to critical scrutiny the fundamental ideas informing current theories of language. Although the focus is on con­ ceptual issues, and the primary target modern 'theories of meaning', we shall examine parallel manifestations of underlying conceptions (and mis­ conceptions) in modern linguistics. In spite of passionate disagreements between the various theorists, there are important and seldom considered common presuppositions. Although we shall not disregard the arguments dividing theorists of meaning, we shall concentrate on what unites them. Our approach to this theatre of war will be indirect. One aspect of this strategy is that we shall focus upon just those topics which are introduced in most theories of meaning with the barest of explanation, taken to be altogether perspicuous and treated with non­ chalance. We shall probe the seemingly clear notion of the truth­ conditions of a sentence, which is commonly taken to be the key to any cogent semantic theory. We shall place pressure upon the apparently obvious distinction, within every sentence, between its descriptive content (the state of affairs it describes, its sense) and its force (e.g. whether it asserts or orders something). We shall test the soundness of the supposi­ tion that a language is a system, a calculus consisting of a network of hidden rules tacitly employed whenever we speak or understand what is spoken. And we shall examine whether the question of how it is possible to understand sentences never heard before really is as deep as it is commonly

The philosopher declares that a theory of meaning is a 'theoretical rep­ resentation of a practical ability',24 while the linguists announce 'The goal of a theory of a particular language must be the explication of the abilities and skills involved in the linguistic performance of a fluent native speaker.'25 The philosopher typically explains linguistic performance in terms of the 'psychological reality' of the theory of meaning he constructs: A theory of meaning will ... represent the practical ability possessed by a speaker as consisting in his grasp of a set of propositions; since the speaker derives his understanding of a sentence from the meanings of its component words, these propositions will most naturally form a deductively ·connected system [viz.· an axiomatic theory of meaning]. The knowledge of these propositions that is at­ tributed to a speaker .. . [is] ... implicit knowledge. 26

In parallel fashion, the transformational-generative grammarian argues that 'Grammar is a system of rules and principles that determine the formal and semantic properties of sentences. The grammar is put to use, interacting with other mechanisms of the mind, in speaking and under­ standing language.'27 This convergence of the two enterprises is explicitly acknowledged: Philosophers of a logical bent have tended to start where the theory was [viz. the formal semantics of logical calculi] and work out towards the complications of natural language. Contemporary linguists, with an aim that cannot easily be seen to be different, start with the ordinary and work towards a general theory. If either party is successful, there must be a meeting.28

If the true Philosopher's Stone is at last almost within our reach, if a theory of meaning, once properly constructed, holds within it the key to 23 D. Davidson, 'Truth and Meaning', Synthese 1 7 (1967), p. 304. 24 M. A. E. Dummett, 'What is a Theory of Meaning? (II)' in Truth and Meaning, ed.

G. Evans and]. McDowell (Clarendon Press, Oxford, 1976), p. 69. 25 Fodor and Katz, 'What's wrong with the Philosophy of Language?', p. 277. 2 6 Dummett, 'What is a Theory of Meaning? (II)', p. 70 (our italics). 27 N. Chomsky, Reflections on Language (Fontana, London, 1976), p. 28, our italics. 2. D. Davidson, 'Truth and Meaning', p. 315.

11

Prolegomenon

Prolegomenon

taken to be. In general we shall resist by argument the theorists' habit of frog-marching the neophyte straight to a ceremony of initiation into the full mysteries of the modern science of language. We shall unmask their conceptual conjuring tricks and break the mesmerizing force of their incantations by critical questioning. Our method will be the clarification of concepts, not the amassing of new evidence about phenomena. A second, related, aspect of our strategy is to by-pass controversies about the truth or falsity of various doctrines in theories of meaning and to focus on the logically prior questions of the intelligibility and purpose of salient theses. A presupposition of the identification of the meaning of a sentence with its truth-conditions is that truth is coherently predicated of sentences and that a certain expression is properly characterized as stating a condition for the truth of a given sentence. If the purpose of a semantic

o f what makes good sense from what strikes us now as nonsensical, depends to a great extent on how we have been educated and on the unquestioned framework of ideas which we have been trained to accept. The claim that some group of thinkers has laboured under certain radical misconceptions is greatly strengthened if we can demonstrate that it is intelligible that they should have thought what they did, even if what they thought now seems patently false or even incomprehensible. To this extent conceptual clarification has an essentially historical aspect and thus differs

12

theory is to show how words engage with the world, then it is assumed that there is a coherent notion of a connection between language and reality on the lines envisaged by the theory. Such questions about the coherence of concepts are matched by questions about the intelligibility of questions. If a senselforce distinction is offered to explain how speakers recognize how an utterance is used, then it is presupposed that the question 'How does a speaker recognize how this utterance is used?' invariably makes sense. Or if an explanation is given to resolve the mystery of understanding new sentences, it is presupposed that the questions 'How does a person under­ stand a sentence?' or 'How is it possible to understand a sentence?' are intelligible questions calling for answers. It is a mistake in addressing theories of meaning to rush headlong into attempts at testing their truth or falsity. There is a fundamental disanalogy between the response appro­ priate to a new theory in physics and that appropriate to the great 'dis­ coveries' in theories of meaning. In the first case the primary issue is typically the truth of an hypothesis, whereas in the second it is usually the intelligibility of a thesis. The abundance of available theories of meaning is no guarantee that any comprehensible answer is being given to any intel­ ligible question. It has frequently happened in the history of human intellectual endeavour that much ingenuity and effort has been expended to no avail because of a defective discrimination of sense from nonsense. Our goal is to trace the bounds of sense in a region where many are now prone to be led astray by grotesque conceptual confusions. A third aspect of our strategy is to cultivate and refine an awareness of the most fundamental elements of the modern conception of language and meaning through an historical sketch of its origins and evolution. The differentiation of what is taken for granted from what is questionable, or

13

from most scientific research. Even if the history of science is generally merely of interest to antiquarians, the history of philosophy and knowledge of the evolution of concepts is vital to making sense of what now confronts us in the guise of theories of meaning, just as the full intelligibility of a painting depends on the historical investigation of a tradition to which it belongs.29 By giving rough sketches, sometimes only a bird's-eye view, of the historical origins of certain contemporary concep­ tions, we shall show the relevance of the history of philosophy and linguistics to the appreciation of what is now of concern in theories of meaning, and we shall urge the importance of history for philosophical understanding, whether in formal logic or in the philosophy of language. Our aim is to illuminate the present by opening up appropriate vistas into the past. The spirit informing this book is obviously sceptical and critical. In pursuing the clarification of concepts, we also demonstrate a readiness to demolish large parts of what pass for significant modern intellectual achievements. But our ultimate purpose is not to persuade linguists or philosophers that their theories are false, thereby encouraging them to redouble their efforts and to construct ever more sophisticated and subtle theories. It is rather to suggest that their endeavours are futile because pointless and misconceived. It will, no doubt, appear to them that we are trying to suppress the light, to deflect them from the true path of a science. In fact we are suggesting that what appears to be sunrise is merely a false dawn, that the path they are following with such enthusiasm leads to the wastelands of the intellect where there is only 'dry sterile thunder without rain'.

29 For a more detailed investigation of this point, see G. P. Baker and P. M. S. Hacker, Frege: Logical Excavations (Blackwell, Oxford, 1984), pp. 4ff.

Historical Bearings CHAPTER 1

Historical Bearings

1

The post-Cartesian tradition

Modern philosophy, typically dated from the writings of Bacon in England or Descartes on the continent, has its roots in the scientific revolution of the sixteenth and seventeenth centuries. It is, no doubt, difficult for us to recapture the sense of excitement, illumination and depth which was associated with the advances in the physical sciences pioneered by Kepler and Galileo and culminating in Newton. Nature was, for the first time, revealed to be intelligible to human reason, subject to laws accessible to human thought. The particular form which such laws assumed was mathematical. Astronomy ceased to be merely the inexplicable 'geometry of the heavens' and became a branch of mathematical physics. Physics, no longer earthbound and confined to the investigation of terrestrial motion in the sublunary sphere, expanded into the study of the fundamental laws governing the behaviour of matter throughout the physical universe. These conquests, however, were achieved only at the cost of a new range of pressing questions. The most dramatic advances in physics involved explanations which employed only geometric or mechanical properties of matter (extension, shape, motion, solidity, etc.). And side by side with the new mechanics there evolved a speculative but fruitful corpuscularian theory of matter, which attributed to matter, in its fine as in its gross structure, only the self-same geometric properties relevant to available physical explanations of observable phenomena. This elementary corpuscularian theory was utilized to sketch a rudimentary scientific theory of perception in terms of the impact of invisible corpuscles, emitted by objects, upon our sensory organs, and the consequent agitation of the 'animal spirits' in the nervous system. Not surprisingly these forms of explanation bred an array of metaphysical and ontological doctrines. The true nature of reality, it seemed, was mechanical and mathematical. Con­ sequently a yawning gulf opened between appearance and reality, between

15

the world as it appears to us to be and the world as it is independently of our observations of it. For the world as it appears to us is multicoloured, noisy, odorous, hot or cold, replete with variegated textures and tastes. But the objective world as conceived by the physicists is only a 'silent buzzing' of invisible particles in motion - all the rest is merely the product of the impact of colourless, tasteless, soundless corpuscles upon our sensibility. If the physicists' story were true (and who could gainsay their remark­ able achievements ?) then a host of new and pressing conceptual questions had to be faced. For if the world is not really at all as it appears to us to be, then can we actually come to know it as it really is? And if we can, how is this possible? Can we know anything at all about the objective world with certainty ? And if we can, what are the marks whereby we may distinguish such genuine knowledge from mere appearance? Philosophers were forced to reconstruct our conception of the objective world and of our cognitive faculties in such a way as would make room for what seemed to be the indisputable advances of science. If the dramatic gulf which the new scientific world-vision had opened up between appearance and reality was to be bridged at all, it seemed that there was little option but to begin from appearances and to construct a picture or conception of reality which was in broad agreement with the deliverances of physics. This required various principles, a priori or empiri­ cal, to legitimate the appropriate inferences from how things appear to us to be to how things, surprisingly and hitherto unexpectedly, really are. Hence seventeenth century philosophy displayed a profoundly egocentric, subjective character. For appearances constitute the subjectively 'given', the impact of the world as it is in itself, upon the sensory apparatus of the human perceiver. The result of such impact is the rich array of sensory data, variously called 'impressions' or 'ideas', which furnish the mind of any subject of experience. The analysis of such sensory data, the discovery of the principles of the mind which organize it into our ordinary, but apparently misleading and deceptive, conception of reality must provide the materials whereby the philosopher can simultaneously justify the ways of the mathematicizing God of physics to man, and also plot the nature, extent and limits of possible human knowledge. The picture with which seventeenth and eighteenth century philo­ sophers worked represented the human mind, equipped with various or­ ganizing dispositions or principles, as receiving, in experience, a wide range of ideas. These included ideas of colour, taste, smell, sound and texture, which, according to the mainstream of thought, misrepresented

16

Historical Bearings

Historical Bearings

the nature of the objective reality which gave rise to them (viz. the hypothesized corpuscularian structure of matter and the corpuscularian mechanisms of perception), and also ideas of shape, motion, extension, solidity ( ? ) , which more accurately represented the nature of the objective world. Further ideas arose, it was argued, from apprehension of the workings of the mind itself, e.g. ideas of the 'passions of the soul'. This huge array of distinct ideas furnished the mind with its main (according to the empiricists, its only) materials for human thought. Philosophical reflection on the possible operations of the mind would reveal both the nature of reality, in so far as it is cognizable, the limits of possible knowledge, and the nature and limits of thought and imagination. This picture moulded the questions which philosophers felt called upon to answer as well as the principal strategies to be used in developing such answers. The ideas with which the mind is furnished could be validated, it seemed, to the extent that they could be shown to be derived, immediately or proximately, from experience, and, if so derivable, to the extent that they could be shown accurately to represent things as they really are. Since not all ideas are so derivable, those which fail the test must be fictitious, and attributable to the workings of the imagination, or innate and so part of the native equipment of the mind, or a priori and structural, and hence attributable to the inherent organizing powers of the mind. The seven­ teenth and eighteenth century debates between rationalists and empiricists polarized around these various, seemingly exhaustive, possibilities. And the salient issues debated, e.g. the possibility of non-trivial a priori knowledge, the possibility and extent of certain knowledge, the analysis and validation of fundamental ideas or concepts such as self, substance, space and time, the relation of mind to body, were introduced and 'resolved' within this framework of thought. Ideas (mental representations, sometimes, but not by any means always, pictorially conceived) constituted the interface between the knowing sub­ ject and objective reality.l By analysing ideas into their simple unanalys­ able components, by scrutinizing their combinations into complexes, by examining their derivation from experience and the principles whereby

I Cf. I. Hacking, Why Does Language Matter to Philosophy? (Cambridge University Press, Cambridge, 1975). Of course, in the hands of Berkeley the apparently mind­ independent objective reality was reduced to an array of mind-dependent ideas appropriately shared out between God and other spirits. Hume went further, for in his metaphysical system both the knowing subject and the objective reality known disintegrate into no more than fictions generated by the workings of our faculty of imagination.

17

they may b e variously organized, philosophers tried to resolve all the epistemological and metaphysical problems which confronted them.

2

The conception of language within the post-Cartesian tradition

The problem-setting context of philosophy during the seventeenth and eighteenth centuries was the picture of reality that informed the develop­ ments in the physical sciences. The materials upon which philosophers conceived themselves as required to work were the various ideas in the mind and the principles of organization whereby the mind orders them to achieve whatever knowledge it may have. Within this framework of thought, language had relatively little claim on philosophical attention. Ideas, and their forms of organization, were typically conceived to be altogether language-independent. And thought, conceived as a variety of more or less complex operations upon ideas, was held to be equally language-independent. A language was generally thought to be no more than a vehicle for the communication of ideas. Hobbes remarked that: the most noble and profitable invention of all other, was that of speech, consisting of names or appellations, and their connection; whereby men register their thoughts, recall them when they are past, and also declare them one to another for mutual utility and conversation . . . The general use of speech is to transfer our mental discourse into verbal, or the train of thoughts into a train of words . . 2 .

Thought was generally conceived as a mental transaction with ideas, private to the thinker and in general independent of language. But since ideas are not themselves transferable from one thinker to another, they can be represented by words. Each man, Locke noted, will 'use these sounds as signs of internal conceptions; and . . . make them stand as marks for the ideas within his own mind, whereby they might be made known to others, and the thoughts of men's minds be conveyed from one to another'. 3 This 'telementational'4 conception of a language was the received wisdom of the age. The Port-Royal Logic (The Art of Thinking ( 1662)), the most influential logic test until the middle of the nineteenth century, proclaimed 2 T. Hobbes, Leviathan, ch. IV. 3 J. Locke, Essay Concerning Human Understanding, III, i, 2.

4 Cf. R. Harris, The Language Myth (Duckworth, London, 1981) pp. 9f.; Harris traces this conception back to Aristotle.

Historical Bearings

Historical Bearings

that 'if the reflections which we make on our thoughts referred to ourselves alone, it would suffice to consider them in themselves, without having recourse to words or any other signs. But . . . we are not able to express our thoughts to each other, unless they are accompanied with outward . signs . . . '5 This picture of language has dominated European thought ever since. It is noteworthy that it is incorporated into theoretical linguistics. Saussure, the founder of this modern science, accepted it without questioning. Speech he conceived as following a 'speech circuit' from the brain (mind) of one speaker to audible words emitted, to sounds received, to the brain (mind) of a hearer, and, given a reply, back again:

words of a language and their concatenation into significant sentences were thought of as mirroring the kaleidoscopic arrangement of ideas into judgments by the mind. This primitive picture, of course, required a host of modifications and epicyclical hypotheses to cope with special cases and complexities. In various more or less sophisticated forms, these were elaborated by the major philosophers of the seventeenth and eighteenth centuries. But the fundamental biplanar conception of a language as composed of words corresponding to ideas and of sentences correspond­ ing to concatenations of ideas into judgments remained a constant guiding force shaping the reflections of thinkers throughout the period. The primitive picture underlying the sophisticated embroideries woven by such philosophical theories as well as by their modern correlates, was later to be dubbed 'Augustine's picture of language' by Wittgenstein.7 It consists of the notion that all words are names of entities (whether ma­ terial, mental or Platonic) and that all sentences are descriptions. Given the conception of a language as a public code in which ideas, concepts or thoughts may be encoded for communicative purposes, it may seem puzzling that our august philosophical predecessors should have had any interest in language at all. Berkeley went so far as to insist that 'so long as I confine my thoughts to my own ideas divested of words, I do not see how I can be easily mistaken. The objects I consider, I clearly and ad­ equately know. I cannot be deceived in thinking I have an idea which I have not.'S Though this may appear faintly comical to a sophisticated modern philosopher of language, Berkeley was merely repeating, in empiricist garb, a conception which runs through the Cartesian method: 'We shall never take the false as the true if we only give our assent to things that we perceive [by 'intellectual vision'] clearly and distinctly . . . [M]ost men apply their attention to words rather than things, and this is the cause of their frequently giving their assent to terms which they do not under­ stand'.9 Nor was this conception of the matter demolished by the repudiation of the New Way of Ideas. Frege ( 1 848- 1925), considered by some to be the fountain-head of modern analytical philosophy, was a relentless opponent of psychologism or idealism in logic, language, and metaphysics. He

18

the opening of the circuit is in A's brain, where mental facts (concepts) are associated with representations of the linguistic sounds (sound-images) that are used for their expression. A given concept unlocks a corresponding sound-image in the brain; this purely psychological phenomenon is followed in tum by a physio­ logical process: the brain transmits an impulse corresponding to the image to the organs used in producing sounds. Then the sound waves travel from the mouth of A to the ear of B: a purely physical process. Next, the circuit continues in B, but the order is reversed: from the ear to the brain . . . in the brain, the psychological association of the image with the corresponding concept. 6

Apart from the quaint physiology, and the substitution of 'concept' for 'idea', the primitive picture is identical. Ideas (on the classical conception) are typically given in experience. These may be complex or simple. Complex ideas decompose upon analysis into simple ones. Language encodes ideas for communicative purposes. Words typically correspond to ideas, definable words corresponding to complex ideas, indefinable words to simple unanalysable ideas. The mind, with its various innate associative propensities and its faculty of imagina­ tion, was pictured as a sort of kaleidoscope. The powers and limits of thought were conceived, by the empiricists, as determined by the pos­ sibilities of arrangement, decomposition and recomposition of ideas. The 5 The Art of Thinking, Introduction; we shall refer to this famous work by its customary name - the Port-Royal Logic. We have used the translation of T. S. Baynes, 2nd edn (Sutherland and Knox, Edinburgh, 1 851). • F. de Saussure, Course in General Linguistics, tr. W. Baskin (Fontana, London, 1 974), pp. 1 1 f. Nor has this picture yet exhausted its mesmerizing power, e.g. 'Language enables a speaker to transform configurations of ideas into configurations of sounds, and it enables a listener within his own mind to transform these sounds back into a reasonable facsimile of the ideas with which the speaker began' (Wallace L. Chafe, Meaning and the Structure of Language (University of Chicago Press, Chicago, 1970)).

19

7 Cf. L. Wittgenstein, Philosophical Investigations (Blackwell, Oxford, 1953), §§lff.; for detailed analysis of the Augustinian picture, see G. P. Baker and P. M. S. Hacker, Wittgen­ stein: Understanding and Meaning, (Blackwell, Oxford, 1980), pp. 33ff. • G. Berkeley, A Treatise Concerning the Principles of Human Knowledge, Introduction, §22. • R. Descartes, Principles of Philosophy, part I, Principles XLIII and LXXIV.

Historical Bearings

Historical Bearings

approached the problems of logic from the very different angle of Platonism, holding that a language encoded not ideas, conceived as mental objects, but concepts and thoughts, conceived as mind-independent Platonic entities. Yet he too argued similarly for an emancipation from language: 'if we . . . attend instead to the true nature of thinking, we shall not be able to equate it with speaking. In that case we shall not derive thinking from speaking; thinking will then emerge as that which has priority and we shall not be able to blame thinking for the logical defects . . . in language.' 10 In fact, the concern of these generations of post-Cartesian philosophers with language was largely a consequence of their conception of a language as an interpersonal code for communicating language-independent thoughts or ideas. The reasons they adduced for examining language were two, one negative and one positive. The negative consideration was that language obscures thought, misre­ presents the ideas which constitute the gold-backing of the paper-currency of words and sentences. The Port-Royal Logic argued that 'the necessity which we have for employing outward signs in order to make ourselves understood, causes us to attach our ideas to words, [so] that we often consider the words more than the things. Now this is one of the most common causes of the confusion of our thoughts and discourse.' I I Conse­ quently, a significant part of that book is concerned with words and p ropositions. So too Locke, who treated 'Of Words' throughout the long Book III of the Essay, excused this digression in the 'Epistle to the Reader':

describe this Idol o f the Market-place 13 in order to emancipate us from its sway. In order to penetrate to the true nature of thought, and thence to the true nature of things, we must free ourselves from the trammels of language. This is to be done by examining the workings of our language and its relation to ideas, noting everywhere how we may guard against error. The positive motive for investigating language was superficially in diametric opposition to the negative one. Language is worthy of philo­ sophical investigation because it is a guide to the nature and structure of the ideas or thoughts it represents, and hence, proximately to the nature and structure of what they represent. The Port-Royal Logic prefaced its remarks on language with the observation that 'it is certainly of some use to the end which logic contemplates - that of thinking well- to understand the different uses of the sounds devoted to the expression of our ideas, and which the mind is accustomed to connect so closely with them, that it scarcely ever conceives the one without the other'. 14 Locke similarly re­ marked that 'The ordinary words of language, and our common use of them, would have given us light into the nature of our ideas, if they had been but considered with attention.'15 Given the basic strategy, it may seem curious that these philosophers thought that any merit at all might attach to scrutiny of language. To the extent that the telementational conception of language viewed language as externally related to thought, it is prima facie paradoxical to contend that we can profitably extract philosophical morals about the nature of thought and its relation to reality from scrutiny of a language, when we can by-pass it altogether and attend directly to the ideas underlying it. This embarrassment is superficially covered up by claiming, as the Port-Royal Logic did, that we are so accustomed to use words in communication that language becomes a partial barrier to reflection upon the ideas behind it. Hence by analysis of language, we can use language itself as a guide to the forms of thought:

20

To break in upon the sanctuary of vanity and ignorance will be, I suppose, some service to human understanding; though so few are apt to think they deceive or are deceived in the use of words; or that the language of the sect they are of has any faults in it which ought to be examined or corrected, that I hope I shall be pardoned if I have in the Third Book dwelt long on this subject, and endeavoured to make it so plain, that neither the inveterateness of the mischief, nor the prevalency of the fashion, shall be any excuse for those who will not take care about the meaning of their own words, and will not suffer the significancy of their expressions to be inquired into.12

Similar sentiments are expressed by most philosophers in this period. Philosophical investigations into language are undertaken by way of a prophylactic. The philosopher, in pursuit of quite a different quarry, must 1 0 G. Frege, Posthumous Writings (Blackwell, Oxford, 1979), p. 270. II

12

Port-Royal Logic, part I, ch. xi. Locke, Essay, 'Epistle to the Reader'. -�"''''

21

since this custom [of verbal expression1 is so strong, that even when we think alone, things present themselves to our minds only in connection with the words to which we have been accustomed to have recourse in speaking to others; - it is necessary, in 1 3 F. Bacon, Novum Organum, xliii, lix: 'words plainly force and overrule the understand­ ing, and throw all into confusion', they 'react on the understanding; and this it is that has rendered philosophy and the sciences sophistical and inactive'. 14 Port-Royal Logic, part II, ch. i. IS Locke, Essay, III, viii, 1 .

Historical Bearings

Historical Bearings

logic, to consider ideas in their connection with words, and words in their connec­ tion with ideas. 1 6

characters must show, when they are used in demonstrations, some kind of connection, grouping and order which are also found in the objects, and . . . this is required, if not in the single words - though it were better so - then at least in their union and connection. This order and correspondence at least must be present in all languages, though in different ways . . . [EJven though characters are as such arbitrary, there is still in their application and connection something valid which is not arbitrary; namely, a relationship which exists between them and things, and consequently, definite relations among all the different characters used to express the same things. And this relationship, this connection is the foundation of truth. 2 .

22

A further consideration reinforced this. If our ideas, to the extent that they are clear and distinct, I7 contain no falsehood, but represent in their combinatorial possibilities, the combinatorial possibilities of the entities in reality of which they are ideas, then our language, to the extent that it is clearly constructed, must represent our ideas and hence also represent, mediately, the self-same possible realities which our ideas represent. Though words are arbitrarily annexed to ideas, though a language is an arbitrary, conventional artefact, once given the arbitrary conventions, representation of thought by language is not arbitrary. Hence the Port­ Royal Logic argued when we speak of the signification of words as arbitrary, there is much that is equivocal in the term arbitrary. !tis indeed a thing quite arbitrary whetherwe join a given idea to a certain sound, rather than to another: but the ideas are not arbitrary things, and do not depend upon our fancy, - at all events those which are clear and distinct . . . [A man's reasoning is] not an assemblage of names according to a convention which depends entirely on the fancy of men; but a solid and effective j udgment on the nature of things, through the consideration of certain ideas which IS he had in his mind, and which it has pleased men to represent by certain names.

This doctrine of approximate correspondence between thought and language and between reality and whatever is clear and distinct in thought contained the seeds of much later, specifically nineteenth and twentieth century, philosophical reflection. It was, however, rarely explicitly thought out or exploited. An exception to the rule, however, was Leibniz.19 In contradistinction to the prevailing tradition, he thought that 'if there were no signs, we should never think or conclude anything intelligibly'.20 In his work a doctrine of correspondence between language and what it represents was more thoroughly and self-consciously developed: ••

Port-Royal Logic, Introduction.

.7 According to the Cartesian conception ideas of colour, sound, smell, are not clear and

distinct, and accordingly do not represent correctly the objective qualities which cause them •• Port-Royal Logic, part I, ch. i. •• In very many ways he stood outside the prevailing tradition, being less enmeshed in the confused doctrine of ideas that bedevilled the empiricist tradition. 20 G. W. Leibniz, 'Dialogue on the Connection between Things and Words', in Leibniz: Selections, ed. P. P. Wiener (Scribner's, New York, 1 95 1 ), p. 10. .

23

Though this relationship is obscured in natural languages, Leibniz dreamt of an ideal language, a characteristica universalis, in which the basic signs would represent simple and unanalysable concepts and all complex signs would be defined precisely in terms of the simple ones. The idea of a universal character was not, as such, original. 22 Leibniz acknowledged his debt to Ramon Lull, Athanasius, Kircher, George Delgano and John Wilkins, and he was familiar with Descartes's discussion of the issue. The bare idea of a universal character was a commonplace in the seventeenth century. In its minimally ambitious form it derived from a perceived need for an international language. The programme for such a project had been formulated by Bacon. What Descartes added to this idea was the vision of an analytical language the primitive words of which would correspond to the simple ideas in the human imagination out of which all human thoughts are compounded, and if [this] explanation were generally received, I would dare to hope for a universal language very easy to learn, to speak and to write. The greatest advantage of such a language would be the assistance it would give to men's judgment, representing matters so clearly that it would be almost impossible to go wrong.23

Such a language would provide an exact and structurally perspicuous system of symbolization for the precise expression of all actual and possible scientific knowledge. Numerous seventeenth century writers pursued this will-o' -the-wisp of a universal character. What Leibniz added to existing projects was the idea of a calculus ratiocinator, a calculus of symbols 2.

Ibid . For a detailed discussion of this fascinating episode in European letters, see L. J. Cohen, ' On the Project of a Universal Character', Mind, LXIII (1954), pp. 49-63. 23 Letter to Mersenne, 20. 1 1.1629, from Descartes: Philosophical Letters, ed. and tr. A. J. P. Kenny (Clarendon Press, Oxford, 1970), p. 6. 22

24

Historical Bearings

which would operate on the formulae of the universal character. Because he had a clearer conception of a logical calculus, and a deeper respect for formal logic, Leibniz's remarks on the correspondence between language, thought and reality are deeper than those of his precursors. The ideal language he envisaged would be an 'algebra of thought' in which simple and indefinable ideas are conceived, as it were, as the alphabet of thought, so that by their combinations all possible thoughts would be expressible. Leibniz envisaged representing these simples by numerical values, and imagined devising a set of combinatorial rules which would synthesize into one system the merits of a deductive logic and those of a Cartesian logic of discovery. That nothing came of these seventeenth century visionary projects is hardly surprising. They muddled up inconsistent requirements and failed to distinguish profoundly different projects. Devising an international language that can be easily learnt and widely used in everyday inter­ national transactions is one thing. It needs a much simplified grammar and pared down vocabulary of a flexible, non-specialized, kind. Concocting a taxonomic vocabulary for scientific classification, description and generalization is a different task altogether. It must involve a specialized vocabulary, often of a highly theory-laden kind (as in chemistry), and the idea that there might be one unique such language suitable for all sciences (never mind about other more important purposes) is a dream founded upon the mythology of the unity of the sciences. So too, the vision of a powerful deductive system in which all patterns of proof and deductive reasoning can be represented is one thing. It may be associated with the hope that some sciences can be given an axiomatic representation in the notation of the calculus. But the notion of a formal logic of discovery, in which 'Calculemus' might be our motto, is another thing altogether. The amalgamation of all these into an envisioned characteristica universalis was a pipe-dream bred of confusion. But the differentiation of the elements of the project was, two centuries later, to bear fruit. The idea of a rough correspondence between thought and language remained a muted but persistent theme in the ensuing period. Thomas Reid, in 1774, argued that Language, being the express image of human thought, the analysis of the one must correspond to that of the other . . . The philosophy of grammar, and that of the human understanding, are more nearly allied than is commonly imagined.2' 2. Thomas Reid, 'A Brief Account of Aristotle's Logic', Works, vol. II, ed. Sir William Hamilton (Edinburgh, 1 863), pp. 691£.

Historical Bearings

25

A century later the same picture was sketched explicitly by Boole ( 1 854): though in investigating the laws of signs, a posteriori, the immediate subject of examination is language, with the rules which govern its use; while in making the internal processes of thought the direct object of our inquiry, we appeal in a more immediate way to our personal consciousness, - it will be found that in both cases the results obtained are formally equivalent. Nor could we easily conceive, that the unnumbered tongues and dialects of the earth should have preserved through a long succession of ages so much that is common and universal, were we not assured of the existence of some deep foundation of their agreement in the laws of the mind itself.25

The philosophical study of language was, on this conception, no more than a useful tool for the truly philosophical investigation into the nature, powers and limits of the human mind on the one hand, and into the essential (metaphysical) nature of reality on the other. Apart from that it provided no more than a negative instruction, a prophylactic against errors rooted in a defective use of, or understanding of, words or in the structural defects of language which only imperfectly or confusedly re­ presents thoughts. The idea that language might constitute part of, let alone the whole of, the subject-matter of philosophy was unthought of. No philosophers dreamt of constructing theories of meaning for natural languages, and the contention that such a theory would constitute a new Philosopher's Stone would have appeared positively bizarre.

3

The logical tradition

We have focused primarily on the mainspring of post-Cartesian philo­ sophical reflection, the attendant conception of language and its relation to thought, and the relevance which investigation into language was conceived to have. Did logicians think differently on the matter? To a modern student, educated in contemporary logic, it is a commonplace that logic is concerned with structural features of languages, since it is typically conceived as studying relations of derivability and entailment between sentences. This idea, however, is a distinctively twentieth century one. With a handful of exceptions (e.g. Whateley and De Morgan in the nineteenth century) the post-Cartesian tradition in logic did not think it

25 G. Boole, The Laws of Thought (Dover, New York, 1958), ch. II, §1.

Historical Bearings

Historical Bearings

the business of logic to study a language and its formal structure, but rather to study concepts and judgments. Descartes, apart from his great influence on the Port-Royal Logic (which, though lucid and elegant, made barely any advance) , contributed greatly to the relative demise of logic in the seventeenth ' and eighteenth centuries. He was altogether contemptuous of formal logic (syllogistic). His reasons were fourfold. First, this traditional logic did not satisfy his requirements of a method, viz. 'certain and simple rules, such that if a man observes them accurately, he shall never assume what is false as true, and will never spend his mental powers to no purpose'. 26 Since formal logic can only guarantee the preservation of truth in valid inferences from the premises to the conclusion of an argument, but cannot guarantee the truth of the chosen premises, it does not satisfy Descartes's demand for a correct method of reasoning. Secondly, he claimed that the syllogism commits a petitio, for the premises must 'contain' the conclusion; hence one can learn nothing new from a syllogism inasmuch as one cannot assert the premises unless one already knows the conclusion to be true. The 'logic of the schools' therefore is only a dialectic which teaches us how to present things already known. (This idea continued to be influential right down to Mill's System of Logic.) Thirdly, formal logic is redundant, since if ideas are analysed according to the Cartesian method into their simplest compo­ nents or 'simple natures', then any possibility of error in inference is ruled out. Inference is a matter of intuitive insight into necessary relations between simple natures. Intuition, Descartes thought, is unerring, and any mistake is attributable to a failure in analysis into clear and distinct ideas. Not only are rules of inference redundant, but they are impediments to thought, since nothing can be added to the 'natural light of reason', the capacity of 'logical intuition', without hindering it. Finally, formal logic is concerned with validity, not with truth; it is a logic of exposition, not of discovery. These views were exceedingly influential, stifling formal logic for two centuries. Locke slavishly followed Descartes, arguing that

With the notable exception of Leibniz, whose influence was negligible, no major philosopher made any significant contribution to formal logic be­ tween the time of Descartes and the revival of logic in the mid-nineteenth century. Kant went so far as to declare

26

This way of reasoning [i.e. a syllogism] discovers no new proofs, but is the art of marshalling and arranging the ones we have already . . . . . . A man knows first, and then he is able to prove syllogistically; so that syllogism comes after knowledge and then a man has little or no need of it.27 26 Descartes, Rules for the Direction of the Mind, rule IV. 27 Locke, Essay, IV, xvii, 6.

27

That logic has already . . . proceeded upon this sure path [of a science] is evidenced

by the fact that since Aristotle it has not required to retrace a single step . . . It is remarkable also that to the present day this logic has not been able to advance a single step, and is thus to all appearance a closed and completed body of doctrine. 28

It is noteworthy that when logic did revive, the originators of the logical renaissance were primarily mathematicians, not philosophers. Neverthe­ less, logic continued to be taught through this period of dogmatic slumber, the most influential work being the Port-Royal Logic. The title of that work, The Art of Thinking, indicates its conception of its subject-matter. It opens with the observation that Logic is the art ofdirecting reason aright, in obtaining the knowledge of things, for the instruction both of ourselves and of others. It consists in the reflections which have been made on the four principal operations of the mind: conceiving, judging, reasoning, and disposing. 29

Accordingly, the book is divided into four parts, each of which examines one of the fundamental operations of the mind. The examination of conceiving analyses the materials of thought, viz. ideas, clarifies their origins, their relations to their objects (that of which they are ideas), their division into categories, their classification into simple and complex ideas. From ideas the mind forms judgments, in which it combines ideas and affirms or denies one or the other. The theory ofjudgment is therefore the analysis of the different kinds of judgments or propositions and of their structure. A proposition is composed of two terms, a subject, of which we affirm or deny something, and a predicate which we affirm or deny of the subject. The copula represents the union of the constituent terms into a judgment. The subject of a proposition may be universal (or singular) or particular, and the proposition may be affirmative or negative; hence the authors follow tradition in reducing all propositions to four kinds, universal affirmative or negative, particular affirmative or negative. Since proposi­ tions may themselves be compounded, a large part of Book II concerns the 2. I. Kant, Critique ofPure Reason, B, viii. 2. Port-Royal Logic, Introduction; 'disposing' is the study of method.

29

Historical Bearings

Historical Bearings

various forms of complex propositions. The rules of reasoning are the norms of valid inference from the premises to the conclusion of an argu­ ment. The exposition of syllogistic in Book III of the Port-Royal Logic became the canonical source of the formal theory of logic for most writers of textbooks of logic for almost two centuries. Throughout this period logic was conceived, in nineteenth century jargon, to be a normative science of judgment. It studied the normative laws of thought, i.e. how inference should be conducted, what ought to be concluded from given premises, etc. What validated these normative laws, how they are related to the descriptive laws of human thinking, remained problematic. That the laws of logic are, in some sense, necessary was readily acknowledged. But what explains their necessity, whence its origin and what its nature was an issue around which writers typically skirted rather gingerly. By the mid-nineteenth century, opinions had polarized around two unhappy alternatives, psychologism and Platonism. Psychologism, which assumed various forms, conceived of the laws of thought as determined subjectively, by the nature and structure of the human mind. This in turn might be explained transcendentally, as in the writings of Kantians, or merely empirically, as in the works of the nineteenth century German psychological school. Sir William Hamilton, a neo-Kantian psychologician, expressed his position in a manner that cannot but seem curious and self-contradictory to the modern reader:

thinking. They are universally valid, provided our thinking remains the same. They are necessary, since to think means for us to presuppose them, as long, that is, as they express the essence of our thinking. 31

28

[I]nsofar as a form of thought is necessary, this form must be determined or necessitated by the nature of the thinking subject itself; for if it were deternlined by anything external to the mind, then would it not be a necessary but a merely contingent determination. The first condition, therefore, of the necessity of a form of thought is, that it is subjectively, not objectively determined. 30

A less Kantian form of psychologism conceded that the laws of thought are valid only conditionally, sub specie humanitatis, in virtue of empirical constraints: We cannot help admitting that all the propositions whose contradictories we cannot envisage in thought are only necessary if we presuppose the character of our thought, as definitely given in our experience: they are not absolutely necessary, or necessary in all possible conditions. On this view our logical principles retain their necessity for our thinking, but this necessity is not seen as absolute, but as hypothetical. We cannot help assenting to them - such is the nature of our presentation and 30 Sir William Hamilton, Lectures on Metaphysics and Logic, vol. III, ed. H. L. Mansel and ]. Veitch (Blackwood, Edinburgh & London, 1860), pp. 24f.

This not only left the status of logical necessity in a curiously contingent limbo, but it also failed to explain the relation between the logician's study of the normative laws of thought and the psychologist's study of the empirical laws of thought. Indeed, it failed to explain how it was possible to reason erroneously, viz. contrary to the forms allegedly determined by the nature of the mind. The alternative conception still viewed logic as a normative science, but adamantly denied that it studied the nature of human thinking. Rather, its subject-matter was the objective relations between mind-independent en­ tities, not ideas, but concepts Platonistically conceived, not judgments viewed as mental objects, but contents of judgment or propositions (the Satz-an-sich) conceived as abstract, sempiternal, objects. The most promi­ nent proponent of this view early in the nineteenth century was Bolzano, and its most august later supporter was Frege. Laws of logic, Frege declared, are not laws of 'takings-to-be-true', but absolute laws of truth: If being true is . . . independent of being acknowledged by somebody or other, then the laws of truth are not psychological laws: they are boundary stones set in an eternal foundation, which our thought can overflow, but never displace. It is because of this that they have authority for our thought if it would attain to truth. They do not bear the relation to thought that the laws of grammar bear to language; they do not make explicit the nature of our human thinking and change as it changes.32

This strategy could readiJrexplain the contrast between the normative laws of thought and the psychological laws of thinking, but only at the cost of Platonic mystery-mongering. Psychologism and Platonism seemed the only viable alternatives. With very few exceptions, neither logicians nor philosophers thought that the laws of logic might be rooted in the essentially arbitrary conventions for the use of arbitrary linguistic signs. Nor did they think that logic was concerned, save indirectly and instrumentally, with language or grammar. 31 B. Erdmann, Logik; quoted by E. G. A. Husser!, Logical Investigations, vol. I, tr. J. N. Findlay (Routledge & Kegan Paul, London, 1970), p. 162. 32 G. Frege, The Basic Laws of Arithmetic, vol. 1, tr. and ed. M. Furth (University of California Press, Berkeley, 1 964), Introduction, p. xvi.

30

Historical Bearings 4 The invention of formal calculi

From the middle of the nineteenth century formal logic underwent an unprecedented revolution. The application of generalized mathematical ideas and techniques to the formal representation of patterns of logically valid reasoning led to the development of modern logical calculi, the powers of which go far beyond anything possible in the limited forms of syllogistic. The philosophical interest of the invention of new logical calculi lay not only in the fact that, and the ways in which, forms of reasoning hitherto beyond the sway of logic became tamed, but also in the philosophical implications of the new discoveries. Did the new logic finally uncover the fundamental forms and mechanics of the human mind? Did it inform us about the ultimate logical or metaphysical structure of reality? Could it, for the first time, reveal the true nature of that great mystery, logical necessity? And did it draw back the veil enshrouding the inner nature of a language? It would be no great exaggeration to suggest that the invention of powerful formal calculi has fulfilled a similar role with respect to mainstream twentieth century analytical philosophy as the new physics fulfilled for seventeenth and eighteenth century philosophy. Where the new physics seemed to require a fundamental re-evaluation of our concep­ tions of appearance and reality, of the powers of the mind and the limits of possible knowledge, the new logic seemed to demand an equally funda­ mental reorientation of our conceptions of language and its relation to the reality it depicts, of the forms and mechanics of understanding, and of the limits of possible thought. The inspiration for the new logic came from the generalization of mathematical techniques. Its pioneer was George Boole (1815-64), a mathematician rather than a philosopher, who sought to clarify and generalize syllogistic reasoning in terms of algebraic operations on sets. His fundamental insight was that there could be an algebra of entities that are not numbers at all, since there are important analogies between dis­ j unction (or conjunction) of concepts, and addition (or multiplication) of numbers. Of course, not all the laws of algebra which apply to the calculus of numbers also apply to the logical calculus of concepts. But that re­ volutionary idea had already been mooted by mathematicians themselves within the domain of mathematics. For with the introduction of new mathematical entities, in particular hyper-complex numbers, and the de­ velopment of an algebra of these objects at the hands of Sir William Rowan

Historical Bearings

31

Hamilton, it became evident that not all the laws of algebra applicable to complex numbers could be retained, e.g. the commutativity of multiplica­ tion. This necessitated a new perspective upon algebra, which was a prerequisite for Boole's invention of logical algebra. For in his calculus of classes xn = x (since the intersection of a class with itself is the same class), and there is no satisfactory analogue of division. With great originality and ingenuity Boole gave algebraic representations of the forms of reasoning recognized in traditional logic. In effect he elaborated a theory of truth­ functions, a technique for their presentation in disjunctive normal form, and a mechanical decision-procedure. His achievement provided the foun­ dations for the invention of calculating machines, and, later, for computers. Boole's mathematicization of logic was a decisive step in the develop­ ment of the subject, and was immediately recognized as such. De Morgan, in 1 85 8 , proclaimed that 'As joint attention to logic and mathematics increases, a logic will grow up among the mathematicians . . . This mathematical logic . . . will commend itself to the educated world by showing an actual representation of their form of thought - a represen­ tation the truth of which they recognize - instead of a mutilated and one-sided fragment . . .'33 Fifteen years later Jevons enthusiastically (and prematurely) declared that 'it will probably be allowed that Boole dis­ covered the true and general form of logic, and put the science substan­ tially into the form which it must hold for evermore. He thus effected a reform with which there is hardly anything comparable in the history of logic between his time and the remote age of Aristotle.'34 Boole no more thought that his new calculus was a study of sentential relations or an exploration of the 'depth-grammar' of any language than his predecessors had thought syllogistic to be an investigation of forms of language. He conceived of logic as the algebra of thought, and character­ ized his abstract algebra as a 'cross section' of rational thinking. His declared aim was 'to investigate the fundamental laws of those operations of the mind by which reasoning is performed . . . and upon this function to establish the science of Logic'. 35 His new calculus of logic, far from being an 'analysis' of the forms of any language, was meant to be a step towards a new, logically more perspicuous, language: 'A successful attempt to express logical propositions by symbols, the laws of whose combinations 3 l Augustus De Morgan, 'On the Syllogism: III', in On the Syllogism and Other Logical Writings, ed. P. Heath (Routledge & Kegan Paul, London, 1 966), p. 78n. 34 w. S. Jevons, The Principles of Science, (Dover, New York, 1958), p. 1 13. 3S G. Boole, The Laws of Thought, (Dover, New York, 1958), ch. 1, §1.

33

Historical Bearings

Historical Bearings

should be founded upon the laws of the mental processes which they represent, would, so far, be a step toward a philosophical language.'36 His mathematicization of logic remained firmly wedded to psychologism. Indeed, he thought that his discovery of a logical algebra was not merely the invention of a useful artificial device, a logically convenient (because logically perspicuous) conventional form of representation, but a revela­ tion of the hidden nature of human thought: 'The laws of thought, in all its processes of conception and of reasoning, in all those operations of which language is the expression or instrument, are of the same kind as are the laws of the acknowledged processes of Mathematics.'37 Ordinary languages in effect conceal, in his view, what his new notation reveals, viz. the abstract mathematical forms of human thought. Boole's ideas were enthusiastically taken up and further developed by other logicians such as Jevons, Venn, Huntington, Peirce and Schroder. But mathematical logic came of age with Gottlob Frege (in 1 879), who acted as Newton to Boole's Copernicus. Where Boole conceived of logic as part of mathematics, Frege aimed to demonstrate that the whole of arith­ metic is deducible from logic. Where Boole had generalized algebraic principles and applied them to logic, Frege invoked a more avant-garde branch of mathematics, namely function theory, subsuming syllogistic within a much more general logical system which sought to display all sound patterns of reasoning as theorems derived from a few function­ theoretic axioms. He explicitly referred to Leibniz's idea of a characteri­ stica universalis, and viewed his function-theoretic calculus not as an analysis of natural languages, but as a logically perfect language which, for the restricted purposes of deductive sciences, would replace natural languages. Calling it 'concept-script' (Begriffsschrift), he conceived of it as accurately representing the structures and articulations of contents of possible judgments (propositions) and of the concepts of which they are composed. He thought of judgeable-contents and of concepts as abstract, Platonic, entities rather than as mental ones. Frege's crucial innovation was to repudiate the traditional conception of (the content of) judgment as composed or synthesized from subject and predicate, and to replace it by a conception of judgment as decomposable or analysable into function and argument. He conceived of a mathematical function as a law of correlation of (mathematical) entities, and accepted the mathematicians' differentiation of functions according to the number

o f arguments taken and according to the types of the arguments. His generalization of function theory beyond the confines of mathematical entities depended upon various forms of liberalization of the concept of a function within mathematics which had taken place earlier in the nineteenth century, just as Boole's generalization of algebra had been dependent on the increasing abstraction of algebra within mathematics. (Mathematicians had already recognized as second-level functions those functions which take first-level functions as arguments, e.g. integration and differentiation.) He took the radical step of recognizing any entities whatever (and not just mathematical ones) as arguments and values of functions. Consequently he proposed viewing judgments (conceived as objects) as the values of concepts for arguments, and concepts as functions mapping arguments on to judgments. The apparatus of function theory enabled him to construct the first formal logic of generality. For he analysed expressions of generality ('all', 'some', 'exists') as variable­ binding, variable-indexed second-level functions. These quantifiers he thus represented as second-level concepts taking first-level concepts as arguments and mapping them on to judgeable-contents. With this apparatus he was able to give the first complete formalization of the predicate calculus with identity, and so to master, for the first time in the history of formal logic, the formal presentation of inferences involving multiple generality (e.g. if every number has a successor, then there is no number such that it is larger than every other number). Like his predecessors, Frege conceived the business of logic to be the analysis of relations between judgments. The well-formed formulae of his concept-script, in the initial formulation of his logical system in Begriffs­ schrift, were conceived as standing for (contents of possible) judgments, just as declarative sentences used in assertion stand for what is judged. Since a precondition for the application of function/argument analysis in logic is that the value of a function for an argument be an 'object', Frege was constrained not only to view the content of a judgment as an (abstract) object, but also to view the symbol representing it as a name. Hence the formulae of concept-script, and indeed the corresponding declarative sentences of natural language, were, ab initio, conceived as singular referring expressions. This idea he never relinquished. But in his final formulation of his system of logic, in The Basic Laws of Arithmetic, he distinguished within an expression for a judgment between the sense expressed which he called 'a thought' (in a non-psychological sense) and a reference, viz. a truth-value: truth or falsehood. This distinction between sense and refer­ ence he also applied to the constituent elements of a symbol for a

32

3. G. Boole, The Mathematical Analysis ofLogic, (Oxford, Blackwell, 1948), p. 5. 3 7 Boole, The Laws of Thought, ch. XXII, § 1 1 .

Historical Bearings

Historical Bearings

judgeable-content, i.e. the argument-expression and function-name (concept-word) are likewise held to express senses and refer to referents. His conception of a concept altered correspondingly, since it was now conceived as a function which takes a truth-value for any possible argu­ ment. The sense of a well-formed formula or declarative sentence was conceived as the object of assertion, or more generally, the sense of any expression was conceived as the mode of presenting a reference (viz. as the value of a given function for some argument). The sense of a sentence (a thought) was conceived as the manner in which a truth-value (its refer­ ence) is presented as the value of a function (denoted by its constituent concept-word) for an argument (denoted by its constituent argument­ expression) . Frege's function-theoretic apparatus was the engine of his logical machinery. His most fundamental innovation was to conceive of concepts as first-level functions taking truth-values as their values for any argument, and of quantifiers as second-level functions (concepts) mapping first-level concepts on to truth-values. It was this which enabled him to construct the formal logic of generality. Once he split judgeable-content into the two strata of sense and reference, conceiving of concepts and truth-values as belonging to the domain of references, he felt constrained to apply his function-theoretic apparatus also to the level of senses. Hence he con­ ceived of the thought, as he had earlier conceived of the judgeable-content, as the value of a function (namely the sense of a concept-word) for an argument (the sense of the argument-expression - in the simplest case, of a proper name). The functional analysis was also applied, ab initio, to the logical connectives. The negation-stroke and the condition-stroke were the two primitive connectives in his system, and he conceived of both as literal function-names. Negation was conceived as a unary function (a concept) mapping a judgeable-content on to a judgeable content. Condi­ tionality was conceived as a binary function (a relation) mapping a pair of judgeable-contents on to a judgeable-content. Although his explanations of these logical operators are isomorphic with modem truth-tabular expla­ nations, they are not identical. In his early system these logical constants are not truth-functions, i.e. functions taking truth-values as arguments, but rather functions taking judgeable-contents as arguments (and values). In his mature system, he construed them as functions taking truth-values as their values, but since he then thought that every function had to be defined for any object as argument (and not just truth-values), he had to offer complex explanations stipulating values when objects other than truth or falsehood are their arguments. They were simply two out of a

myriad of concepts and relations, with no more right to be deemed truth-functions than any other concept or relation. 38 This difference between Frege's conception and that of a modem logi­ cian is not a mere slip of the pen, but rather reflects his conception of logic and its laws. His explanations of the logical constants are not conceived as formal definitions, and the axioms of his logical system are not presented as consequences of these explanations, but as indemonstrable self-evident truths. The axioms of logic are no more analytic consequences of defini­ tions than are the axioms of geometry (on his view); they are rather an unfolding of the essences of primitive concepts and relations. Just as our faculty of spatial intuition delivers the axioms of geometry, so too the axioms of logic are directly certified by a 'logical source of knowledge'. It is these axioms which secure the certainty of all the validly derived theorems of logic, but they are ultimate intuitive truths, not verbal trivialities. They can no more be justified than the ultimate truths of geometry: 'The question why and with what right we acknowledge a law of logic to be true, logic can answer only by reducing it to another law of logic. Where that is not possible, logic can give no answer.'39 This conception is a reaffirmation of the venerable idea that all knowledge ultimately rests on self-evident truths. It leaves the normative status of logical truths (as canons of correct reasoning) shrouded in clouds of Platonic mysteries. In the course of his exposition of his logical system, Frege introduced numerous novel ideas which provided grist for philosophical and logical mills for subsequent decades. Apart from those already mentioned we shall select three more which proved seminal. First, he propounded as a salient principle of his analysis that a word has a meaning or content (signifies something) only in the context of a sentence. This principle is evidently associated with his insistence upon the priority of judgments over concepts for purposes of logical analysis, viz. that we view concepts as derived by functional decomposition of judgments, rather than viewing judgments as synthesized from antecedently given concepts (subject and predicate). That principle in tum can be seen as a straightforward consequence of his function-theoretic analysis. For al­ though functions are, in his view, abstract, language-independent entities, the logical type of a function is determined by the categories of its argu­ ments and values. Since, in his initial system, the judgment is the value of a function (concept) for an argument, the function (concept) is viewed as

34

35

3. For a more detailed and precise statement of this daim, see G. P. Baker and P. M. s. Hacker, Frege: Logical Excavations, (Blackwell, Oxford, 1984), pp. 1 17f. 39 Frege, The Basic Laws ofArithmetic, vol. 1, xvi.

Historical Bearings

Historical Bearings

abstracted from the judgment by considering how arguments (in the simplest case, objects) are correlated with values (viz. judgments). Frege employed the principle for a variety of purposes. In particular, he insisted that in concept-script the sense of an expression that is a constituent of a well-formed formula consists in its contribution to determining the truth­ value which is the reference of the formula.40 His successors seized on his 'contextual principle' and enshrined it as the foundation stone of modem philosophical semantics, viewing the sentence as taking priority over its subsentential constituents in the semantic analysis of language. Secondly, although Frege agreed with the venerable tradition that all valid inferences are from judgments (assertions) to judgments (assertions), he emphatically distinguished (as did many of his predecessors) between the judgment or assertion and what is judged or asserted. That which is asserted (the content of judgment, or 'thought') can occur unasserted, e.g. in the antecedent of a conditional. This is the basis of the claim that he distinguished between force and sense. A sentence or formula may be uttered with assertoric force, but the sense it expresses (viz. that such-and­ such) remains constant even if it is not asserted but occurs as the antece­ dent of a conditional. He did not generalize this distinction to imperatives, optatives and other syntactical forms. But it has become a major inspira­ tion of modern philosophical semantics. Thirdly, Frege carried further the old idea of a form of correspondence between language, thought and reality. His concept-script; conceived as a characteristica universalis41 as well as a calculus ratiocinator, and designed according to the canons of function theory, displayed perspicuously the logical articulations of the thoughts it expressed and the metaphysical

articulations of the objects and concepts it denoted. Natural language only corresponds in a rough and ready fashion with the forms of thought and the metaphysical structure of reality. Language is not constructed from a logical blueprint, Frege insisted, and one can no more learn logic from studying grammar than one can learn how to think from a child. Natural language, he thought, is rife with vagueness, ambiguity, lack of logical perspicuity, and, indeed, logical incoherence. To a large degree he identified as 'logical defects' in a language those features of it which fail to correspond with the articulations of his concept-script. The logical powers of concept-script in the presentation of arguments so far outstripped anything hitherto available that Frege unwittingly employed his invention as a yardstick against which to measure natural language. The old idea of isomorphism between a form of representation and thought or reality was, however, resuscitated by the invention of a much improved logical calculus. This idea too was to bear fruit in the next stage in this history. It is noteworthy that the invention of powerful formal calculi gave impetus to the 'biplanar' or Augustinian conception of language. It is a distinctive feature of Frege's calculus, conceived as an improvement over natural language for special scientific purposes, that in it all the constituent symbols (with the exception of variables and the assertion-sign) stand for extra-notational entities. To be sure, he distinguished entities into differ­ ent logical types (objects, functions, first-level concepts, second-level concepts, etc.) and differentiated kinds of names according to the different kinds of entities they stood for (proper names, first-level concept-words, second-level concept-words, etc.) . He also correlated each name not with one entity, but with two, viz. a sense and a reference, the former being an abstract entity which determined the latter as correlated with the name. But it was precisely the sophisticated function-theoretic apparatus inform­ ing these complex innovations which also preserved the fundamental Augustinian picture, since the core idea of a function is that of a law of correlation of entities. It took only one step from Frege to conceive of a formal calculus as a perfected syntax of a language merely awaiting an interpretation, which would be given by assigning entities in reality to the uninterpreted signs of the calculus. This became even clearer in the reflections of Russell, who pursued further Frege's vision of reducing arithmetic to logic, and who developed further the formal apparatus of a rich function-theoretic calculus:42

36

40 Cf. Frege, The Basic Laws ofArithmetic, vol. 1, §32. It is on the strength of this passage that Frege is generally held to be propounding a truth-conditional theory of meaning. This is, however, highly misleading. What he claimed was that by stipulating references for con­ stituent names of his formulae, he also thereby assigned a sense to a well-formed formula, namely that given such-and-such references of constituent names, the conditions under which the formula (sentence) refers to the truth-value True are fixed. The thought expressed by such a formula is therefore the thought that these conditions are fulfilled, viz. that the True is the value of the denoted function for the denoted argument. Since we do not conceive of sentences as names of truth-values, nor of truth-values as objects that are the values of functions for arguments, this conception of sense (the mode of presentation of a truth-value as the value of a given function for a specified argument) cannot coherently be said to encapsulate the modern idea that the sense of a sentence is given by specification of its truth-conditions. 41 Though rid of all the seventeenth century associations between a universal character and an international language useful for commerce and easily learnable by all, as well as the idea of a single unified taxonomic language for the whole of science.

37

42 The following illustrative quotation postdates Russell's association with Wittgenstein, and bears the mark of the latter's influence. Russell's conception in Principia was very different.

38

Historical Bearings

In a logically perfect language the words in a proposition would correspond one by one with the components of the corresponding fact, with the exception of such words as 'or', 'not', 'if', 'then', which have a different function. In a logically perfect language, there will be one word and no more for every simple object, and everything that is not simple will be expressed by a combinat.ion of words, by a combination derived, of course, from the words for the simple things that enter in, one word for each simple component. A language of that sort will be completely analytic, and will show at a glance the logical structure of the facts asserted or denied. The language which is set forth in Principia Mathematica is intended to be a language of that sort. It is a language which has only syntax and no vocabulary whatsoever. Barring the omission of a vocabulary I maintain that it is quite a nice language. It aims at being that sort of language that, if you add a vocabulary, would be a logically perfect language.43

Russell's famous Theory of Descriptions, which showed how to paraphrase sentences containing singular referring expressions of the form 'The so­ and-so' into sentences which do not, was designed to demonstrate that the appearance in ordinary language of significant phrases lacking a reference, like 'the present King of France' or 'the golden mountain', only decep­ tively suggested that there could be meaningful expressions which nevertheless did not stand for something. The true logical form of a sentence of the form 'The cper tps' is 'There exists one and only one object such that it both cps and tps', and here each significant expression, in the forms of the language ofPrincipia, actually stands for something. Thus the Theory of Descriptions was designed to preserve intact the putative in­ sights of the model of correlation enshrined in the Augustinian picture. The conception of a logical calculus as an ideal, and ideally perspicuous, syntax to which semantics adds an interpretation by assigning entities to the constituent designating symbols was to dominate subsequent philo­ sophical reflections. Unlike Russell, Frege had little interest in drawing any extensive general philosophical morals from his logical investigations save in the area of philosophy of mathematics. Indeed, it was here (and not in formal or indeed philosophical logic) that the raison d'etre of his endeavours lay. For the whole point and purpose of his labours in inventing a concept-script lay in devising a proof of the reducibility of arithmetic to logic. Apart from this, which proved a failure, he had no interest in exploring in anything but minimal detail any implications within philosophy of the possibility of representing the forms of thought and inference in the notation of a formal 43 B. Russell, 'The Philosophy of Logical Atomism', in Bertrand Russell: Logic and Knowledge, essays 1 901 -50, ed. R. C. Marsh (Allen & Unwin, London, 1956), pp. 1 97f.

Historical Bearings

39

function-theoretic calculus. This task was left to his successors, in particu­ lar the young Wittgenstein.

5

The watershed

Wittgenstein differed from his predecessors above all in conceiving of logic neither as the study of language-independent relations between abstract entities nor as the study of the laws of thinking, but rather as the investiga­ tion of the fundamental forms of any system of symbolic representation whatever. One great question dominated his first book, the Tractatus Logico-Philosophicus: what are the most general conditions for the possi­ bility of representation? The abstract logical structure of a correct con­ ceptual notation would reveal, he thought, the essential nature of any possible sign-system which can represent reality. Hence, unlike his pre­ decessors, he thought that ordinary language was in good logical order. 44 For to constitute a system of representation at all, it must have that essential structure which makes representation possible, viz. logical structure. However, that structure is not perspicuously revealed in the surface forms of a natural language, but is only uncovered by philosophi­ cal analysis. Hence investigations into an ideal logical notation are not a project of devising an 'improvement' relative to any natural language (as Russell, Frege, and their predecessors had thought) but rather a matter of bringing to light what is hidden in the symbolism of a language itself. The essence of representation, Wittgenstein thought, lies in description, in the representation of a state of affairs by means of a proposition. Whether such a description is used to assert that that is actually how things are in reality, or whether it is used to query whether things are thus, or to order that things be made thus, is a further matter which is of no concern to logic. All that interests logic is the unasserted proposition. Since we describe how things are in reality by means of propositions (conceived neither as abstract entities nor as mental ones but rather as sentences with a sense), the simplest possible proposition must contain whatever essential features are required for description (representation). Hence the beginning of wisdom must lie in clarification of the logical nature of the 'elementary proposition', viz. a sensible sentence which is 44 This is not to say that it is without defects. Nor is it to say that it contains no pseudo-propositions. It is rather that what it does say, it says clearly. There is no halfway house between expressing a sense, and being a nonsensical pseudo-proposition. The genuine propositions of ordinary language are in good order.

40

Historical Bearings

logically independent of every other such sentence, i.e. which has no (non-trivial) entailments, but merely describes the existence of an elementary state of affairs. Investigating the essence of the elementary proposition (which is the first part of Wittgenstein's renowned 'picture theory of the proposition') will give the essence of all description, and that in turn will capture the essence of the world. For what is essential for a description is whatever makes possible the representation of states of affairs in reality, and whatever structural features states of affairs possess in virtue of which they can be described at all are themselves the essence (the essential form) of reality. Wittgenstein's logical investigations were, therefore, elaborated on a vast backcloth of the most general philosophical import. For, in his view, the investigation into the essence of the proposi­ tion will reveal the essential nature and limits of language, the range of all possible worlds and the limits of thought, since thought too is representa­ tion. 'My work', he wrote, 'has extended from the laws of logic to the essence of the world'. 45 Wittgenstein had welcomed the demise of traditional subject/predicate logic, and greeted function-theoretic logic with enthusiasm, although also with fundamental disagreement. Frege and Russell had pioneered the route to the Promised Land, but had not arrived there. Like Frege and Russell, he conceived of a proposition as a function of its constituent expressions,46 i.e. as decomposing into function and argument. He ac­ cepted a version of Frege's sense/reference distinction, but quarrelled with Frege's employment of it. He conceived of sentences as having senses, but denied that they had references. Fully analysed names had references, but no senses. An elementary proposition, he argued, is composed of simple indefinable names. These denote simple sempiternal objects in reality, which were conceived as the metaphysical substance of the world (such as spatio-temporal points; simple, unanalysable qualities). The connection of these names with their referents is arbitrary and conventional. We combine these names to form elementary propositions according to conventional logico-syntactical rules which specify possible (significant) combinatorial possibilities. These conventions determine the logical forms of the names. To assign a given object to an arbitrary name as its referent, the logical 45 L . Wittgenstein, Notebooks 1 914- 1 6, 1st edn, ed. G. H. von Wright and G. E. M. Anscombe (Blackwell, Oxford, 1 96 1 ), p. 79. 46 Wittgenstein, Tractatus, 3.31 8 ; Frege had conceived of a judgeable-content as the value of its constituent concept, Russell of the proposition as the value of its constituent proposi­ tional function.

Historical Bearings

41

syntax o f the name must mirror the combinatorial possibilities of the object in reality, viz. its ability to combine with other such objects to constitute a state of affairs (e.g. a given spatio-temporal point's having such-and-such a quality). Though the names of objects are arbitrary, and though logical syntax is conventional, once these conventions are fixed by us, the relation between an elementary proposition and the state of affairs it depicts is essential and internal. The proposition must be isomorphic with the state of affairs it depicts, it must have the same logical multiplicity and identical logical form. These theses enshrined a form of the Augustinian picture at the heart of the Tractatus, as a picture, not of the surface forms of language, but of its underlying structure. It is of the essence of an elementary proposition to be bi-polar, i.e. to fix a 'logical space' which reality either does or does not fill. Contrary to Frege, truth and falsity are not accidental features of a sentence with a sense, but rather to have a meaning or sense is to be either-true-or-false, i.e. to restrict reality to a 'Yes/No possibility'. The combination of names according to the rules of logical syntax represents, is a logical picture of, a state of affairs. Once we know the meanings of the unanalysable names, and grasp their mode of combination, we know what states of affairs are represented by their configurations in elementary sentences. To under­ stand a proposition all that is necessary is to understand its constituents (names and form); what we then know is what is the case if it is true (although not necessarily whether it is true). It is this, Wittgenstein thought, that explains the fact that we can understand sentences we have never heard before. The rules of the logical calculus of language determine in advance all possible significant sentences, and mirror all possible elementary states of affairs in reality. This doctrine of strict isomorphism between fully analysed sentences and reality, of an internal, necessary, relation between the sentence and what it depicts which is a consequence of arbitrary conventions, was indeed novel. But the depth of the gulf separating Wittgenstein's philosophy from that of his predecessors only emerges clearly in his account of molecular propositions, of the nature of their mode of combi­ nation, and of his consequent radical conception of logic, logical truth and necessity. The Grundgedanke (fundamental thought) of the Tractatus, he an­ nounced, was that the logical constants (logical connectives, quantifiers, identity sign) are not representatives. Frege had conceived of these crucial syncategorematic expressions as denoting actual concepts and relations (unary and binary functions), mapping (inter alia) truth-values on to

43

Historical Bearings

Historical Bearings

truth-values. Wittgenstein thought this was quite wrong:47 these expres­ sions are not names at all, and there are no such 'logical objects'. Rather, they signify 'operations' whereby we generate compound propositions from elementary ones. Any atomic proposition may be true and may be false. We can draw up truth-tables correlating any number of propositions with all possible distributions amongst them of truth-possibilities. Coordi­ nation of propositions with 'logical constants' simply indicates a select distribution of truth-possibilities amongst the constituent propositions. Thus 'p & q' is true if and only if both p and q are true; 'p v q ' is true if either 'p' or 'q' is true or both are true, etc. And Wittgenstein showed how the different combinations thus selected can be generated by the operation of joint negation upon sets of propositions. The logical constants are thus defined by appropriate truth-tables. The truth-grounds, or truth­ conditions, of a molecular proposition consist of those distributions of truth-possibilities of its constituent propositions that make it true. The sense of a molecular proposition is thus given by its truth­ conditions, which reality may or may not satisfy. Thus 'p & q' allow reality only the conjoint realization of the states of affairs p and q, whereas 'p v q' allows the realization of p alone, of q alone or of both. It is a cardinal thesis of the Tractatus that all significant propositions are analys­ able into truth-functional combinations of atomic propositions; quantified propOSItions are conceived to be infinite conjunctions or dis­ junctions of propositions, and apparently non-extensional contexts (such as 'It is necessarily true that p' and 'A believes p') are shown by analysis either to be pseudo-propositions or to be, in fact, extensional.

Typically the truth-value of a molecular proposition will depend on how truth-values are distributed among its constituents. But there are two limiting cases, viz. a tautology, which is true however truth-values are assigned to its constituents, and a contradiction, which is false however such as assignment is made. Tautologies and contradictions are the prop­ ositions of logic, e.g. - (p & -p), ((p ::> q) & p) ::> q. They are limiting cases of propositions, for they are not bi-polar: tautologies leave open, as it were, the whole of logical space, since they are true however things are in the world, and contradictions close off the whole of logical space, since they are false no matter what. The propositions of logic have no sense, for they have no truth-conditions, tautologies being unconditionally true, contradictions unconditionally false. All true propositions of logic, Wittgenstein thought, are tautologies. And all, uniformly, have no content. They do not, as Frege thought, describe fundamental timeless relations between abstract entities, nor are they, as Russell believed, a description of the most general features of the universe. They are simply (senseless) consequences of our conventional symbolism for combining propositions:

42

47 He gave various reasons, e.g. (i) the so-called 'primitive signs' of logic '�', ' v ' , , : ', '&', are interdefinable, hence not genuine primitive signs, nor names of relations; (ii) a function cannot be its own argument (one cannot substitute the function 'g is a fish' in the argument place of 'g is a fish', but one can take the result of an operation, e.g. '�p' and use it as the base for that very operation to produce '��p'; (iii) operations can 'vanish', e.g. ��p p, � (3x) �fx (x)fx, (3x)fx. (x a) fa; (iv) If'�' were a function name, then '��p' would be a different proposition from 'p'; (v) If ,� , were a genuine function, infinitely many proposi­ tions would follow from every elementary proposition, viz. '��p', '����p', etc., which is absurd. He also pointed out that Frege's explanation of negation as � function which takes the true as its valUe for the false as argument (and the false as value for any other object as argument) leaves the sense of this function absolutely undetermined, despite these stipula­ tions. For on this account, provided 'p' and 'q ' have the same truth-value, e.g., false, then '�p' and ' �q' would have the sam!! sense, viz. that the false falls under the concept of negation, which is absurd. His reasoning was extended, with additional arguments, to encompass identity and quantification. )

=

=

=

=

not. In some things are arbitrary in the symbols that we use and . . . some things are which in field a not is logic it is only the latter that express: but that means that logic nature the which in one rather but signs, we express what we wish with the help of of the natural and inevitable signs speaks for itself . . . ng the One can calculate whether a proposition belongs to logic, by calculati logical properties of the symbol. without And this is what we do when we 'prove' a logical proposition. For, out of ion proposit logical the t construc we , bothering about sense or meaning signs.'8 with deal that rules only others by using

This radically conventionalist conception of logical truth dispersed the clouds of psychologistic and Platonist mysteries that had always en­ shrouded the notion of logical necessity. 49 Logical necessity, Wittgenstein thought, was merely a consequence of our arbitrary conventions of logical syntax for the compounding of propositions. And the only genuine (express48 Wittgenstein, Tractatus, 6.124, 6.126. of necessary truth. 49 Wittgenstein was not the first to propose a conventionalist account

hi­ Hobbes, in De Corpore, and Berkeley, in his private notebooks later published as Philosop cal Commentaries, had similar inklings. But neither developed or attempted to demonsttate this thesis.

44

Historical Bearings

ible) necessity is logical. 50 His bold thesis of analysis (viz. that every proposition has a unique analysis into truth-functional combinations of elementary propositions), his truth-tabular definitions of the logical con­ nectives and his demonstration that the truth-tables provide a mechanical decision-procedure for the propositional calculus, led to a major reorien­ tation of the conception of logic. Unlike Frege, Wittgenstein insisted that one can draw inferences from false, unasserted, propositions. He objected to Frege's presentation of logic as an axiomatic system resting on a privileged set of self-evident axioms. 'All the propositions of logic are of equal status: it is not the case that some of them are essentially primitive propositions and others essen­ tially derived propositions.'51 All the propositions of logic follow from operations generating tautologies out of tautologies; which is selected as the starting point is entirely arbitrary. Whether a proposition is a taut­ ology is settled conclusively by the decision-procedure of the truth-table method. Proof in logic, in which a logical tmth is 'derived' from premises which ultimately rest on axioms, is merely an expedient for recognizing complicated tautologies, which can be done by direct inspection of truth­ tables. Logic is not a science resting on self-evident axioms, but the manipulation of signs determined by rules of logical syntax. 'If we know the logical syntax of any sign-language, then we have already been given all the propositions of logic'. 52 Logic has no grounds. It must, Wittgenstein insisted, 'take care of itself'. It is not only logic, but philosophy itself that emerges from the furnaces of the Tractatus in a new mould. The analysis of the underlying structure of language brought to light the nature of logical necessity. It also purported to clarify the essential limits of what can be said in a language. The cardinal propositions of philosophy, conceived since the dawn of the subject as revealing ultimate truths about the nature of reality, the metaphysical structure of the world, transpire to be ill-formed proposi­ tions which violate the rules of logical syntax. They are not bi-polar, since they are conceived as necessary truths about reality. But only bi-polar propositions picture reality; and only tautologies are necessary truths, and 50 Metaphysical truths such as abound in the Tractatus itself were clearly conceived as necessary, but, according to Wittgenstein, could not be expressed in language save by nonsensical pseudo-propositions that violate logical syntax. He never relinquished his idea that all necessity is conventional, but later came to deny that it is always explicable in terms of truth-functional composition. 51 Wittgenstein, Tractatus, 6. 127. 52 Ibid., Tractatus, 6. 124.

Historical Bearings

45

they say nothing about the world. The typical philosophical 'propositions' employ illegitimate categorial concepts (substance, property, concept, etc.) as if they were genuine names, but analysis reveals them to be variables, not names. Consequently these metaphysical pronouncements are no more than pseudo-propositions. What task then remains for philosophy? Philosophy is a critique of language. By logical analysis of language it makes clear the legitimate empirical propositions correctly expressed in the confusing garb of the surface grammar of a natural language, and it demonstrates the illegitimacy, the nonsensicality, of puta­ tive philosophical propositions. There are, and can be, no genuine philosophical propositions. Philosophy is not a body of doctrine, but an activity of clarification. From this highly compressed sketch of some of the main theses of the Tractatus it is worth extracting a list of fundamental principles most of which were to play a pivotal role in subsequent developments in philosophy: (i) Ordinary language is in good logical order. (ii) The essential function of a language is depiction (description) : the representation of a state of affairs, which may or may not be realized. (iii) Assertions, questions, commands, optations contain a descriptive com­ ponent (the unasserted proposition, later called 'the propositional­ radical') . It alone, the differences being relegated to psychology, is of concern to logic. (iv) Every proposition has a unique analysis. The surface grammar of a sentence conceals its logical form; logical analysis reveals its under-

lying structure. (v) All elementary propositions are logically independent of each other. They are the last (propositional) product of analysis. it (vi) The elementary proposition is isomorphic with the state of affairs depicts, and is internally related to it. gs (vii) The sense of an elementary proposition is a function of (the meanin of) its constituent names and its mode of composition. (viii) Knowledge of the meanings of constituents of propositions (names ition. and forms) suffices for understanding the sense of any propos ation explan fresh This is how it is possible to understand without sentences never encountered before. (ix) Every compound proposition is a truth-function of elementary propositions (thesis of extensionality). Hence every meaningful proposition is constructible by a series of operations on elementary propositions.

46

Historical Bearings

(x) The sense of a compound proposition is given by its truth-conditions, i.e. the expression of agreement and disagreement with the truth­ possibilities of its constituent elementary propositions. (xi) All entailments are consequences of inner complexity of propositions. One proposition entails another if the truth-grounds of the first include those of the second. Logical analysis will lay bare the logical powers of propositions. (xii) A language is, au fond, a calculus of signs. A correct function­ theoretic logical calculus reveals the essential structure of any possible language. (xiii) Speakers of a language possess tacit knowledge of the underlying forms and elements of a language which are concealed by surface grammar. Enormously complicated tacit conventions are pre­ supposed by everyday speech. (xiv) All necessity is logical necessity; logical necessity is a consequence of our conventions for the use of signs requisite for any system of representation by a symbolism. (xv) Philosophy is the activity of clarification of a language by logical analysis. Wittgenstein's successors, first the logical positivists and then theorists of meaning, accepted at least a version of most (though not all) of these principles. Wittgenstein himself was later to repudiate most (though not all) of them. Ironically, those he retained tended to be precisely those which philosophers, toiling along the paths plotted out by the Tractatus, repudiated. Like Captain Nolan at Balaclava, Wittgenstein tried to warn his followers from galloping headlong in the direction which he had wrongly pointed out to them. And like Nolan's, his warnings went unheeded.

CHAPTER 2

Sense and Force: the Evolution of the Species

1

The background environment

Inspired by the Tractatus and taking encouragement from the writings of Tarski and Camap, troops of philosophers have charged off in the direc­ tion of elaborating theories of meaning for natural languages on the foundation of the principle that the meaning of a sentence is its truth­ conditions. They conceive of such a theory as a finite array of axioms (the definitions of the primitive terms of the language) and a finite set of principles (specifications of how each of the formation- and transformation­ rules which suffice to deliver all well-formed type-sentences contributes to determining the meaning of a complex expression from the meanings of its parts). Together these are thought to constitute a deductive system the theorems of which give the meanings of the well-formed sentences by stating, in a canonical form, their truth-conditions. Such a theory is also held to specify the meaning of any subsentential expression (including the primitives) by clarifying its contribution to the truth-conditions of any sentence in which it may occur. The fundamental achievement is advertised as a demonstration of how the meaning of a sentence depends on its structure and the meanings of its constituent words, for this makes trans­ parent 'how finite resources suffice to explain the infinite semantic capacities of language'. · Philosophers have been reinforced in this crusade by legions of theoreti­ cal linguists. They labour in the faith that a language is a system (or a system of systems). The task of anything worthy of the label 'a semantic theory' must be to expose general features of well-formed expressions I D . Davidson, 'Semantics for Natural Languages', in Linguaggi nella Societa e nella Tecnica, ed. B. Visentini (Edizioni di Comunita, Milan, 1970), p. 177.

48

Sense and Force

which have systematic effects on the determination of the meanings of those expressions. Many linguists apparently consider that the best hope to achieve such a theory lies in exploiting the identification of the meaning of a sentence with its truth-conditions. On this basis a flourishing branch of linguistics seeks to 'account for the meaning of each expression on the basis of a patterned exhibition of a finite number of features'. 2 The effort and enthusiasm lavished by philosophers and linguists on truth-conditional semantic theories is most remarkable. So too is the convergence of ideas and purposes between these two very different groups of theorists. To anybody who conceives of the construction of a semantic theory as an enterprise au fond scientific, the convergence itself fosters the conviction that both packs must be hot on the scent of the truth. Surely, it seems, the framework of investigation must be sound, and any residual conundrums will ultimately be removed by progressive sophisti­ cation and refinement of accepted concepts and principles. On the other hand, the bystander must immediately be struck by the precariousness of the theorists' position. Indeed, they seem to be threatened with a complete rout once battle is joined. The reason for this is simple. The meaning of any expression is taken to be its contribution to the truth-conditions of any sentence in which it occurs. But many sentences are not expressions of anything that can be evaluated in the dimension of truth and falsity. It would be nonsense to declare that what the sentences 'When was the battle of Hastings fought?' or 'Turn out the light before you go to bed !' express is true or false. Yet, if such sentences cannot be characterized as true or false, would it not follow that it is nonsense to speak of their truth-conditions? More well-formed sentences of a natural language apparently fall outside the scope of such a theory of meaning than fall within it. Consequently, as a general theory of meaning, it is a non-starter. It must be declared bankrupt before it has even started trad­ ing. Could anybody seriously contend that questions, requests, and com­ mands have no meanings ? Or that their constituent words are deprived of meanings in these sentence-frames ? Would it be any better to claim that individual words have different meanings in non-assertoric utterances? In the face of this threat, some protective reaction seems required. The standard manoeuvre is to introduce a distinction between the sense of a sentence and its force. Different theorists draw such a distinction in different ways, or more accurately, they draw many different distinctions among which there is a complex set of interrelationships. But they are 2 Ibid., p. 1 77.

The Evolution of the Species

49

guided by a number of principles which generate kinship among the forms of the various senselforce distinctions. To a first approximation, a sensei force distinction distinguishes within type-sentences components, fea­ tures, or aspects which are bearers of sense and force and which meet the following constraints: (i) Every type-sentence has both a sense and a force, i.e. both a 'sense­ conveying component' and a 'force-indicator'. (ii) The force of a sentence determines (at least in part) what speech-act is performed by an utterance of this sentence, e.g. whether a question is asked, an assertion made, or an hypothesis aired. (iii) Only a complete type-sentence has a force. (A half-sentence cannot be used to perform any such speech-act.) (iv) There is a high degree of freedom in combining senses with forces. Almost any sense can be attached to any given force, and almost any force can be linked to a given sense. In particular, any sense what­ ever can be endowed with assertoric force. (Equivalently, there is a great latitude in combining 'sense-conveying components' with 'force-indicators'.) (v) The sense of a sentence is a bearer of truth-values. Hence it can be ascribed truth-conditions. The force of a sentence has no connection with truth and falsity, and therefore it has no bearing on the specifi­ cation of truth-conditions. (Accordingly, a 'force-indicator' must lack sense and a 'sense-conveying component' must lack force.) The role of a senseiforce distinction is to relate the uses of non-assertoric sentences systematically to entities (the senses of sentences) which are exhaustively characterized by reference to truth-conditions. A theory of force would embroider on the following simple model: a speaker under­ stands the sentence 'Is it raining?' if he knows that it is typically uttered to ask a question and if he can identify that what is asked is whether the proposition that it is raining (viz. the same proposition which is asserted in the sentence 'It is raining') is true. A viable distinction between sense and force would justify a division of labour. Truth-conditional semantics will be confined to specifying the senses of sentences, leaving the clarification of any principles concerning force to another science (pragmatics). Provided that every meaningful sentence, whether or not what it expresses is open to assessment as true or false, is systematically related to something which bears truth-values and hence can intelligibly be held to have truth-conditions, and provided that this transformation does not map sentences obviously different in meaning

Sense and Force

The Evolution of the Species

on to the same entity, then the possibility emerges of arguing that truth­ conditional semantics does offer a complete theory of meaning for the sentences of a language. This possibility would be realized, it seems, by the construction of a theory which specified how systematically to determine the force of a sentence and how systematically to relate its force and its truth-conditions to its actual use in communication.3 Such a theory could then be applied to block derivation of the theses that non-assertoric utterances have no meaning and that words differ in sense in assertoric and non-assertoric utterances. Protected by unlimited drawing rights on an hypothetical theory of force, truth-conditional semantics, apparently sol­ vent again, is suddenly back in business. Without this support it would apparently have to wind up. No wonder a philosopher declares 'It is difficult to see how a systematic theory of meaning for a language is possible without acknowledging the distinction between sense and force, or one closely similar.'4 The sense/force distinction is widely held to be a sine qua non of truth-conditional semantics, hence indirectly of the analysis of the logical structure of language and the essential nature of human understanding. 'We should not have the least idea,' it is alleged, 'how such a theory of meaning might be constructed if we were not familiar with the distinction, introduced by Frege, between sense and force. '5 The idea has an independent appeal to theoretical linguists. They can elaborate the notion of force into a theory of pragmatic performance,

The sense/force distinction thus offers a fresh justification for a traditional division of labour within theoretical linguistics. Distinguishing sense from force is not merely an artificial manoeuvre promising strategic advantages, but rather an idea enjoying strong intui­ tive support. Two independent lines of thought converge on it. One takes rise from nomenclature of traditional European grammar. Different moods are differentiated in the conjugation of verbs: indicative, impera­ tive, subjunctive, and (in some languages) optative. Also, sentences are classified into different forms: declarative, interrogative, and imperative. These terms indicate relations between the forms or constituents of sen­ tences and their standard uses. A simple declarative sentence whose verb is indicative is typically used to make an assertion, an interrogative sentence to ask a question, and an imperative to issue an order or request. In the standard classification of verb-forms and sentence-structures we seem to be already familiar with the rudiments of a theory of force. Of course, the correspondence between grammatical characterizations and the uses of sentences is only rough and approximate (e.g. an interrogative sentence may be used on occasion to make a statement, i.e. as a 'rhetorical ques­ tion'), but modern linguists and philosophers might well be able to refine these intuitions into a precise theory for each particular language. It is also noteworthy that traditional grammar adumbrates systematic relations between different sentence-forms. Common school exercises concentrate on transforming declarative sentences into corresponding sentence­ questions, positive imperatives into corresponding negative ones, etc. Standard classifications are linked with a simple network of rough cor­ respondences, which once again look like raw materials for building rigorous theories. Moods and basic sentence-forms provide a gateway into a theory of force. The second source of inspiration for differentiating sense from force is reflection on the patterns of reports in indirect speech. If a speaker addresses the sentence 'What time is it?' to a passer-by, we would record what he did by saying 'He asked what time it was'; if he responds to a request for a weather-forecast by saying 'It will rain', we might make the report 'He predicted that it would rain', etc. Four features of such reports are noteworthy. First, there are more or less systematic methods for generat­ ing statements in oratio obliqua, methods which are often taught and learned in language-classes. Secondly, reports in indirect speech fall into two parts: the specification (more or less precise) of what act the speaker performed by uttering a sentence (e.g. 'he asserted that . . . ' or 'he promised that . . .') and the identification of the object of this act (e.g. 'What he

50

VIZ.

a theory about how speakers and hearers figure out the utterance-meaning of sentences used in non-null contexts on the basis of their knowledge of grammatical principles and information about the contexts . . . The conception of a theory of pragmatic performance is introduced only to provide a theoretically motivated distinction between the domains of semantics and pragmatics that will enable us to keep questions about the use of sentences out of semantics.6

3 M. A. E. Dummett, 'What is a Theory of Meaning? (II)', in Truth and Meaning, ed. G. Evans and). McDowell (Oxford University Press, Oxford, 1976), p. 74. 4 M. A. E. Dummett, Truth and Other Enigmas (Duckwotth, London, 1978), p. 450. 5 M. A. E. Dummett, 'What is a Theory of Meaning?', in Mind and Language, ed. S. Guttenplan (Clarendon Press, Oxford, 1975), p. 72. • ). J. Katz, Propositional Structure and Il/ocutionary Force: A Study of the Contribution ofSentence Meaning to Speech Acts (Harvester, Sussex, 1977), p. xiii; but some linguists, e.g. G. Gazdar, follow philosophers in defining pragmatics simply as what lies outside truth­ conditional semantics.

51

52

Sense and Force

The Evolution of the Species

asserted was that nitrogen liquifies at - 1 79 °C' or 'What he asked was what time it was'). The main verb typically indicates the act performed, the

between grammatical aspects of sentences and the typical uses of these sentences. Even more impressive, however, is the fact that differences in mood and sentence-forms can be directly matched in sentences exploiting the resources of indirect speech. Instead of uttering the sentence 'Hydro­ gen is lighter than oxygen' to make an assertion, I could instead utter the sentence 'I assert that hydrogen is lighter than oxygen'; instead of saying 'What time is it?' to ask the time, I might say 'I wonder what time it is' or 'I would like to know what time it is'; and instead of saying 'Shut the door' to give somebody an order, I might say 'I order you to shut the door'. In the case of each of those paraphrases, the corresponding report of the utter­ ance in indirect speech exactly matches the structure of the sentence uttered, and hence it seems that each of these paraphrases contains a phrase ('I assert . . .', 'I wonder . . .', 'I order . . .') which makes explicit the use to which the utterance is actually put. The possibility of transforming sentences into ones of this general form seems to put the senselforce distinction beyond question. For, in each paraphrase, the prefix appar­ ently indicates what the force is (and nothing but the force), while the indirect statement, question, or command states what the sense is (and nothing but the sense). Clarifying how the distinction between sense and force is to be drawn for sentences of a given natural language will appar­ ently be tantamount to explaining how to determine whether particular paraphrases of this form are correct or not. Everything seems to point to the possibility and importance of distinguishing sense from force and of applying this distinction in constructing a theory of meaning for a language. It is altogether unsurprising to find that part of the standard procedure of theorists of meaning is to do just this. Indeed, so standard has this become that the performance is fully ritualized. Frequently no explanation at all is supplied of how the terms 'sense' and 'force' (or their equivalents) are to be employed. It is supposedly enough for a theorist to announce that he intends 'to analyze all sentences, declarative or non-declarative, into two components: a sentence-radical that specifies a state of affairs and a mood that determines whether the speaker is declaring that the state of affairs holds, commanding that it hold, asking whether it holds, or what'. 7 Every competent theorist is presumed to understand and to accept the intended distinction; no major uncertainty is envisaged, and no doubts are canvassed or countered. Apparently there is universal agreement about a

indirect statement (indirect question, etc.) the object of this act. Thirdly, this articulation makes perspicuous the facts that different speech-acts (e.g. assertion, conjecture, and supposal) may have the same object and that the same act may have different objects. There are two independent dimensions in which to describe what speakers do by uttering sentences, and this gives rise to multiple possibilities for describing connections between utterances. Finally, indirect statements are the typical grammati­ cal subjects of predications of truth and falsity. If, e.g. I utter the sentence 'Herodotus understood ancient Persian', then what I have asserted (viz. that Herodotus understood ancient Persian) is false. Consequently, the expression specifying the object of an assertion, prediction, conjecture, etc. is apparently the designation of a bearer of truth-values. These four features of indirect speech mesh with some of the requirements for a theory built on distinguishing sense from force. Paraphrasing utterances into oratio obliqua makes explicit what speech-act the uttered sentence was used to perform, and at the same time it offers a systematic way of extracting from the uttered sentence another expression (the correspond­ ing indirect statement) which expresses something true or false and thus is capable of being characterized by its truth-conditions. The indirect state­ ment can therefore be regarded as a formulation of the sense of the sentence whose utterance is reported in oratio obliqua. And since the identification of the speech-act performed in uttering this sentence de­ pends on features of the uttered sentence together with relevant aspects of the particular circumstances of its utterance, the force of the sentence can be identified with the typical use of sentences displaying these features. The route from utterances to reports in oratio obliqua traces out a general contour-line the exact delineation of which seems to be the proper business of a theory of sense and a complementary theory of force. The support that each of these separate considerations affords to dis­ tinguishing sense from force is redoubled, it seems, by the observation that the two lines of reasoning can be dovetailed together. The features of sen­ tences which determine jointly with the context what are the correct reports of the speech-acts performed by particular utterances are plausibly j ust those familiar aspects of sentences distinguished in traditional gram­ matical classifications of sentence-forms and the moods of verbs. Hence somebody who introduced a senselforce distinction on the basis of oratio obliqua reports would be led to attend to the same fundamental features of sentences as somebody who began from the traditional correspondences

53

7 David Lewis, 'General Semantics', in Semantics of Natural Language, 2nd edn, ed. D. Davidson and G. Harman (Reidel, Dordrecht, 1972), pp. 20S£.

Sense and Force

The Evolution of the Species

unique distinction between sense and force. This underpins a clear division of labour between semantics and pragmatics, and it transforms a seem­ ingly insuperable objection to truth-conditional semantics into a vindica­ tion of this general framework for investigating the notion of meaning.

introduced to represent assertoric force. I I All these authors suggest that a general correspondence holds between grammatical distinctions and diffe­ rent uses of sentences, and all three follow tradition in relating the different uses of sentences to differences in mental acts, states, or attitudes. Frege is widely thought to have made a decisive advance on this conven­ tional wisdom. A distinction between sense and force is commonly cata­ logued among his major achievements, and he is universally hailed by modern theorists as the ultimate source of inspiration for their own senselforce distinctions. What support is there for these dramatic claims? His rather perfunctory remarks about sentence-forms in natural languages hardly seem to merit any accolade. Others had noticed rough correspond­ ences between uttering declarative sentences and making assertions, be­ tween uttering interrogative sentences and asking questions, etc. And so too had they suggested that different speech-acts manifested different relations of speakers to common objects (ideas, propositions, contents). Where was Frege's originality? It is all tied up with his concept-script. He introduced a special symbol (the 'judgment-stroke' or 'assertion-sign') to express the act of judging something to be true. 12 In the conviction that the clear representation of this act is essential to judging the cogency of an inference in concept-script, he prefaced this symbol to every separate stage of a proof, i.e. to every premise, to every intermediate conclusion, and finally to the proposition to be proved. Hence every line of a proof in concept-script has the general form

54

2

Phylogenesis

A distinction of sense from force can be used to turn the tables against the most obvious objection to the thesis that the meaning of a sentence is its truth-conditions. The argument, which we have just sketched, to a consider­ able degree recapitulates the evolution of the modern versions of this distinction. The idea that the typical use of a sentence corresponds to such conspicuous features of it as the mood of the main verb and its syntactic form is an ancient one among grammarians, logicians, and philosophers. For example, the Port-Royal Grammar suggested that the different moods of the verb signify different 'movements of the soul' (affirmation, desire, and com­ mand) ; the indicative was thought to manifest the attitude of judgment towards thoughts, while the subjunctive, optative, and imperative were held to manifest different volitive attitudes.8 The authoritative status accorded to the works of the Port-Royal school guaranteed that these doctrines remained part of the conventional wisdom of educated men down to very modern times. Frege shows the influence of these ideas in his convictions that the assertoric force of an utterance is bound up with the indicative mood of the verb9 and that the assertoric utterance of a sentence is the outward counterpart of a psychological act of judging a thought to be true. So too does Russell in his claim that the three sentences 'Beggars are riders', 'Are beggars riders ?', and 'Beggars shall be riders' each present the same proposition, adding that in the first this is asserted, in the second presented as the object of a doubt and in the third as the object of a volition. 10 Even Wittgenstein accepted a similar doctrine at the time of the Tractatus, for he characterized the difference between judgments (asser­ tions), commands, and questions as merely psychological, and hence he denied any proper place in logic to Frege's judgment-stroke which was Grammaire Gem!raie et Raisonnee, chs XIII and XVI. 9 G. Frege, Posthumous Writings, ed. H. Hermes et al. (Blackwell, Oxford, 1979), pp. 2, •

1 29, 1 39, 149, 1 94, 1 98, 252; 'The Thought', in Philosophical Logic, ed. P. F. Strawson (Oxford University Press, Oxford, 1 967), p. 22. 1 0 B. Russell, 'Theory of Knowledge' (unpublished rypescript, 1 913), p. 196.

55

�A Here the vertical bar signals the act of judgment, while the rest of the formula (the 'content-stroke' or 'horizontal' and the other symbols) merely expresses the proposition, thought, or idea which is judged to be true. It is this articulation of Frege's symbolism which has so inspired later theorists. Although he introduced his notation in Begriffsschrift with an explana­ tion that mentioned his intention to symbolize the act of judgment and 'judgeable-contents', he later altered his terminology and referred to 'as­ sertoric force' and 'thoughts' (i.e. 'senses of sentences'). Apparently he 1 1 L . Wittgenstein, Notebooks 1 91 4- 1 6, ed. G. H. von Wright and G. E. M. Anscombe

(Blackwell, Oxford, 1961), p. 96; Tractatus, 4.023, 4.442. 1 2 G . Frege, Begriffsschrift, tr. T. W. Bynum, in Conceptual Notation and related articles (Clarendon Press, Oxford, 1972), §2; The Basic Laws of Arithmetic, vol. 1, tr. and ed. M. Furth (University of California Press, Berkeley, 1964), §5.

Sense and Force

The Evolution of the Species

held that the addition of assertoric force is necessary to make the expres­ sion of a thought into an assertion, and therefore the 'judgment-stroke' could be viewed indifferently as symbolizing assertoric force or the act of assertion. Since the rest of a well-formed line of a proof in concept-script is the mere expression of a thought, a distinction beMeen sense and as­ sertoric force is transparent in the articulation of his symbolism. Frege also stressed that these two elements are independent of each other. Any thought expressed with assertoric force can also be expressed without it, and any true thought expressed without assertoric force may legitimately be combined with this force. In particular, he harped on the thesis that the antecedent of a conditional expresses a thought without assertoric force. This case he cited as conclusive proof that any thought may be expressed without being asserted to be true. Finally, he distinguished the judgment­ stroke from another symbol in concept-script which expressed a different act, viz. the ' double-stroke of definition I t- '. This can be prefixed only to a well-formed formula whose principal logical constant is the identity-sign; it is not used to express a judgment, but rather to lay down a rule that a newly introduced symbol is to have the same content or sense as an already understood symbol. 13 Hence he apparently included a contrasting pair of force-indicators in his concept-script as well as symbolizing the difference between presence and absence of assertoric force. The excitement derived from these explanations is enhanced by some important asides. One is a late observation about interrogative sentences. Frege declared that a declarative sentence and the corresponding sentence­ question 'contain the same thought'; they differ in that the first 'contains an assertion', while the second 'contains a request'. 14 He did not, however, suggest that imperative or optative sentences express the same thought as corresponding declaratives, indeed he explicitly denied that they express thoughts at all. The other noteworthy observation is that the introduction of the judgment-stroke into concept-script is grounded in the 'dissociation of assertoric force from the predicate'. I S This has the ring of a decisive repudiation of the venerable dogma that the indicative mood of the verb or the copula is what effects the act of assertion. And this in turn makes possible a definitive dissolution of philosophical puzzles about deductive inference. It seems, for example, that modus ponens involves a petitio, since the conclusion that q is already asserted in the premise that if p then q

(hence independently of knowing whether it is true that pl. By distinguish­ ing assertion from predication, Frege allegedly cleared up this confusion. Predication is relevant to sense alone, while assertion is a matter of force. This whole case for honouring Frege as the founder of what is now understood to be a senselforce distinction rests on misinterpretations of his writings and on turning a blind eye to defects and confusions in his thinking. His own explanation of his formal symbolism is arguably in­ coherent and certainly sabotages the purpose for which we might wish to use it. In the background of his exposition are two mistakes. One is the claim that inference is possible only from judgments or assertions, i.e. from thoughts actually advanced as true. This is the ultimate rationale for the need to symbolize the act of assertion (or assertoric force) in concept­ script. And it is, by modern lights, an obvious blunder. The second misconception is that the predicate of a declarative sentence is what carries assertoric force. Since he deliberately abandoned subject/predicate structure in the symbolization of judgments in his concept-script, he had to invent a special symbol to serve as the vehicle of assertoric force; otherwise any formula in his notation would necessarily lack assertoric force, and hence it would be inadmissible in any cogent proof. 'Dissociation of assertoric force from the predicate', far from repudiating it, actually reaffirms the doctrine traditional in earlier grammar and logic. If either of these misapprehensions had been corrected, Frege's entire case for intro­ ducing the judgment-stroke would have collapsed. Moreover, without a commitment to a quasi-mechanical conception of assertoric force, even conceding that inferences must hold among assertions need not justify his notation; his purpose would have been served as well by the stipulation that every independent line in a proof is to be taken to have assertoric force. The more immediate purpose of Frege's novel symbolism diverges from the rationale for modern senselforce distinctions. He intended to mark in his notation the distinction between a judgment (a judgeable-content or thought judged to be true) and a mere judgeable-content (or mere thought). 1 6 But his point was not to contrast the use of a complete sentence to make an assertion with the use of the same (or another) complete sentence to perform another speech-act (e.g. formulating a conjecture or framing a supposition), but rather to contrast the use of an expression as a complete sentence to make an assertion with the use of the same expres­ sion as a clause in a compound sentence (e.g. a conditional or disjunction)

56

13

Frege, Begriffsschrift, §23 ; The Basic Laws of Arithmetic, vol. 1, §27. Frege, 'The Thought', p. 21. 15 Frege, Posthumous Writings, pp. 1 85, 198. 14

16 Ibid., pp. l In., 1 85, 1 98.

57

Sense and Force

The Evolution of the Species

where it is not used to perform any speech-act at all (or at least none coordinate with making an assertion, asking a question, etc.). These are altogether different issues. In the first case, the alternative to having assertoric force is having some other force (e.g. imperative or interroga­ tive), whereas in the second it is having no force at all. The second case can in fact be dealt with by a single completely general principle, viz. it makes sense to ask what speech-act is performed by uttering an expression (or what force this expression has) only if the expression uttered is a complete sentence (i.e. not part of some more complex sentence). It may be im­ portant to note that no clause ever is used to make an assertion - important at least for avoiding philosophical confusions. But this simple point can be established without any general investigation of speech-acts. That strategy was the one that Frege followed. He strove to make clear that a thought may occur unasserted as part of a thought which is asserted. In such cases, he held (misleadingly) that a speaker 'expresses a thought' without judging it to be true. Relatively late he cited an actor's declaiming on the stage and someone's uttering a sentence-question as further instances of expressing a thought. But his purpose was not to investigate these speech-acts; it was to prove that 'it is possible to express the thought without laying it down as true'. 1 7 In fact he conflated two distinct issues under the label 'to express a thought' (below, pp. 101-2), and hence this putative proof is fallacious. More detailed inspection of his writings renders even more absurd the contention that Frege introduced a sense/force distinction along modern lines. His suggestion that sentence-questions express the same thoughts as the corresponding declarative sentences was strictly limited. He did not claim that word-questions express thoughts, and he denied that impera­ tives and optatives do. This rules out any possibility of using his distinction of sense from force to encompass all sentence-forms within a framework of analysis based on truth-conditions. Moreover, he refrained from refer­ ring to interrogative force. 18 His conception of sentence-questions has clear precedents, and in any case it is not even consistent with his own analysis of oratio obliqua (because the sense expressed by a declarative sentence is designated by an indirect statement while that expressed by a sentence-question is designated by an indirect question; but a correspond­ ing such pair of expressions are not intersubstitutable salva signi(icatione, let alone salva veritate, hence they cannot be co-designative). Finally, his symbolism for concept-script does not exploit the possibility of contrast-

ing assertoric force with any other. Although the assertion-sign has an apparent foil in the double-stroke of definition, the latter cannot be construed as attaching some non-assertoric force to a thought for the simple reason that the formula to which it is prefixed does not yet have any sense at all. Almost none of the distinctive principles informing modern sense/force distinctions are satisfied by Frege's representation of sense and force. Russell and Wittgenstein moved somewhat closer to modern forms of a senselforce distinction. Russell envisaged that a common 'proposition' could be filtered out of appropriately related declarative, interrogative, imperative, and optative sentences. This 'proposition' he held, might be expressed in a form denuded of force, viz. a verbal noun. So the proposi­ tion common to 'The cat is on the mat', 'Is the cat on the mat?', 'Would that the cat were on the mat' is expressed by 'The eat's being on the mat'. He did not develop this curious idea. His purpose was eliminative: he wished to brush aside those aspects of sentences which do not bear on the act of grasping a proposition which alone is the proper subject-matter of logical investigation. Wittgenstein had a parallel negative aim: he banished the distinction between assertions, commands, and questions from the domain of logic in order to focus exclusively on the 'unasserted proposition'. Neither had a theory of force. 19 Both in effect declared force to be irrelevant to logic, and both left in the dark the question of how to settle whether different sentences (or utterances) expressed the same proposition. It was the growth of truth-conditional semantics which supplied as it were the genetic impulse favourable to the development of modern ver­ sions of a sense/force distinction. In the hands of the logical positivists, the basic principle of truth-conditional semantics was transformed into the principle of verifiability: the meaning of a sentence is the method or conditions of its verification. Non-declarative sentences were pushed to the sidelines as cognitively meaningless, and so too were many declarative sentences. Pronouncements in metaphysics, religion, and morals were widely condemned. Moral utterances were likened to imperatives or to ejaculations of approval and disapproval (e.g. 'Up with honesty' or 'Murder - ugh !'), and they were held to be beyond the pale of logic and rati9nal deliberations (so that 'practical reasoning' was considered a misnomer). The influence of logical positivism was widespread immediately before and after the Second World War.

58

17

Frege, 'The Thought', p. 21.

1 . Still less did he broach the issue of the 'force' ofthe lines spoken by an actor on the stage!

59

19 Any more than Frege had a theory of modality (d. Begriffsschrift, §4); indeed their attitude to force resembles his attitude to modality.

Sense and Force

The Evolution of the Species

The modern notion of force first made a conspicuous appearance in writings critical of this form of truth-conditional semantics. The leading figure of this reaction was J. L. Austin. As early as 1 946 he suggested that philosophers, under the influence of logical positivism, begin many investi­ gations of concepts by committing the 'descriptive fallacy'. 20 Epistemo­ logists go astray in this way. They consider that a sentence such as 'I know that it is a goldfinch', because of its declarative form, must be used to make a statement; they then enquire into what would make it true, thereby treating it as a description of the speaker's performance of 'a specially striking feat of cognition, superior, [but] in the same scale as believing and being sure'. 2 1 According to Austin's diagnosis, this generated nonsense. The expression 'I know', like 'I promise', is really used, he argued, not to describe or report anything, but to do something, or to perform an act. Hence he dubbed these expressions 'performatives' and sentences contain­ ing them 'performative utterances'. He originally argued that such utter­ ances cannot be assessed in the dimension of truth and falsity at all and that they lack truth-conditions because they are not descriptions. There­ fore a proper account of what they mean, how they are used, and what constitutes a successful performance of these speech-acts (the 'felicity­ conditions' of promising, ordering, etc.) lies beyond the scope of truth­ conditional semantics. A full account of the utterances of a natural language requires a further supplementary theory for non-descriptive utterances. This will give a philosophical analysis of what it is for an utterance to have the force of a promise, order, request, greeting, etc. The variety of forces that utterances may have, i.e. the variety of distinct speech-acts that they may be used to perform, is enormous, whereas truth-conditional semantics is capable of clarifying only those expressions which are used to make statements or give descriptions. Austin soon expanded his critical remarks into the outline of a novel positive theory22 which was meant to embrace both performative utter­ ances and assertions (or descriptions). He suggested that every utterance (and its constituents) has a more or less definite sense (and reference), and in addition that it is used to perform some particular 'illocutionary act' (e.g. asking or answering a question, giving a warning, announcing a verdict). Alternatively, he said that every utterance has a particular illocutionary force, which is the speech-act actually performed by its

production. If uttered with the appropriate intention in the proper circumstances, 'There is a bull in the field' may have the illocutionary force of a warning, or in a different situation, the force of a description (of the landscape). On this view, force is a feature of particular utterances, not of type-sentences, and there are as many distinct forces as there are different speech-acts. Although Austin ascribed both sense and force to every utterance, he left wholly unexplained what the sense of a sentence is. In particular, he gave no hint about how to specify the sense of a Wh-question or an imperative, and consequently it is unclear whether he thought that every sense could be expressed with assertoric force or that every force could be combined with any given sense. He apparently aimed to produce a general theory which would analyse assertion as a species of the genus of speech-acts and which would exhibit truth and falsity as merely one out of a range of equally important dimensions for assessing the 'happiness' of utterances. But he did not live to work out such a theory in any detail. A parallel movement of thought leading in a different direction is visible in the work of R. M. Hare in moral philosophy. He took exception to the framework of analysing moral utterances advocated by logical positivism. The primary role of such typical moral pronouncements as 'A person ought to keep his word' or 'It was an evil act of yours to . . .' is not to describe anything, but rather to evaluate something, e.g. to commend it, to condemn it, or to prescribe the doing of some deed. Consequently, accord­ ing to The Language of Morals, the proper method for clarifying the meanings of such expressions as 'good', 'ought', and 'right', is to focus on their role in bestowing commendatory, prescriptive or evaluative force on utterances rather than on the loose and variable descriptive content which they have. 'Good' is declared to have an 'evaluative meaning [which] is constant for every class of object for which the word is used', and whatever descriptive meaning it bears in a particular utterance is 'secondary to the evaluative meaning'.23 Hare urged making 'a distinction between the meaning of the word "good" and the criteria for its application';24 the latter determine its variable descriptive content, but the former fixes its constant force of commendation. Logical positivism went off on the wrong track because it failed to acknowledge the fundamental importance of the observation that to call something 'good' is to commend it. Once again the initial thrust of this reasoning is that truth-conditional semantics is too narrow in scope to embrace an analysis of the meaning of all utterances (in this case, of moral utterances). 23 R.

60

20 J. L. Austin, 'Other Minds', Philosophical Papers, 1st edn (Clarendon Press, Oxford, 1 96 1 ) , p. 71. 21 Ibid., p. 67. 22 J. L. Austin, How to Do Things with Words (Clarendon Press, Oxford, 1962).

M. Hare, The Language of Morals (Clarendon Press, Oxford, 1952), p. 1 18. 24 Ibid., p. 102.

61

Sense and Force

The Evolution ofthe Species

Hare moved on from this negative point in the direction of building a more general positive theory very different from Austin's. His purpose was to develop the analysis of the speech-act of commending far enough to show that logical positivism was mistaken to deny the possibility of practical reasoning and so to impugn the rationality· of morals. Hare carried out this programme by accepting in a modified form the identifica­ tion of moral judgments with commands which had been suggested by some logical positivists. He then urged that there is a logic of imperative inference parallel to the logic of assertoric inference. The raw materials for this are what he called 'neustics' and 'phrastics'.25 Imperative and indica­ tive sentences may, he suggested, be about the same thing. They differ in such a case in that one commands the bringing about of a state of affairs, whereas the other describes it. 'Shut the door !' and 'You are going to shut the door' are both alleged to be about your shutting the door shortly. The verbal noun 'your shutting the door in the future' indicates what both are about, and this expression Hare called the 'phrastic'. The connection between the order and the assertion is then made transparent by so paraphrasing both that they share this phrastic. Hare transformed the command into 'Your shutting the door - please' and the assettion into ' Your shutting the door - yes'. The expressions 'please' and 'yes', as used in these paraphrases, he called 'neustics'. The neustic marks the speech-act performed by uttering the corresponding unanalysed sentence, in particu­ lar differentiating commands from assertions. On this slender basis he gave an account of imperative inference and thus, indirectly, of practical reasoning. The logic of imperatives simply boils down to the logic of phrastics together with the principle that at least one premise must have imperative force if the conclusion does. Or, more accurately, Hare aimed 'to reconstruct the ordinary sentential calculus in terms of phrastics only, and then apply it to indicatives and imperatives alike'. 26 The Language of Morals did not use the apparatus of neustics and phrastics to construct a general theory of meaning. No neustics other than 'yes' and 'please' are considered, and no claim is made that all speech-acts can be analysed into assertion and command (e.g. asking a question). Moreover, Hare's purpose is served by the comparatively weak thesis that every command has a counterpart assertion (not requiring the converse). But in response to criticisms of his analysis of moral utterances, he later moved further in the direction of a general theory, adding 'tropics' to his

equipment of 'neustics' and 'phrastics', generalizing to the claim that every utterance has a phrastic, neustic, and tropic, and analysing questions into imperatives (commands to fill in blanks or to assign truth-values to sentences). 27 A flood of theorizing has sprung forth from the work of Austin and Hare. For various different reasons theorists have sought to prove that every utterance or sentence can be analysed into a neustic and a phrastic, a force-indicator and a descriptive content, and recently the main motiva­ tion for this enterprise has been to show that truth-conditional semantics, by explaining descriptive content, provides the foundation, when supple­ mented by a theory of force, for a complete explanation of how language works or of how communication is possible. We will review two branches of this torrent of work in philosophy and theoretical linguistics. The first directly focuses on logic. It is now believed that Frege and Wittgenstein introduced a purely semantic conception of validity for inferences involving assertions: an argument is valid provided that the satisfaction of the truth-conditions of the conjunction of its premises guarantees the satisfaction of the truth-conditions of its conclusion. Since this explanation has no immediate application to sentences not used to make assertions, it becomes a moot point whether a distinction between valid and invalid can be drawn for non-assertoric reasoning and also whether such reasoning, assuming its possibility, obeys the same logical principles as assertoric inference. A distinction between sense and force is now commonly exploited to demonstrate the possibility of valid non­ assertoric reasoning and to work out the basic principles of the 'logic of imperatives' and the 'logic of questions'. Emulating Frege's use of the assertion-sign to mark assertoric force and his practice of attaching this to the mere expression of a propositional-content, modern philosophers introduce symbols for other forces, e.g. ' ! ' to mark commands and ' ? ' to mark sentence-questions. They produce such formulae as '?(p /\ q) ' or ' ! (p � q) ' , and they propose such rules of inference28 as

62

! (p !p

26

Ibid., pp. 1 7ff. Ibid., p. 26.

q)

!p I- (q � p ) !q

and

?L ?rp

27 R. M. Hare, 'Meaning and Speech Acts', Philosophical Review, LXXIX (1 970), pp. 3ff.

28

25

v

63

Cf. A. J. P. Kenny, 'Practical Inference', Analysis, 26 ( 1 966), pp. 72ff,; A. ]. P. Kenny, Will, Freedom. and Power (Blackwell, Oxford, 1 975), p. 70; F. Waismann, Principles of Linguistic Philosophy, (Macmillan, London, 1 965), p. 405.

64

Sense and Force

The Evolution of the Species

Their general strategy is to relate codifications of intuitively acceptable non-assertoric reasoning systematically to the relations among the truth­ conditions of the unasserted propositional-contents. In this way they aim to vindicate the thesis that the underlying principles of valid reasoning are everywhere the same, superficial appearances to the contrary notwith­ standing. Separating sense from force is indispensable to carrying out this project. The second stream of work tries to use a distinction between sense and force or neustic and phrastic as the gateway into a general theory of meaning. This strategy was explicitly pioneered by E. Stenius in his study of the Tractatus. His motive was to argue that Wittgenstein's Philosophi­ cal Investigations was not intended to be a root-and-branch repudiation of the Tractatus, but rather a modified and defensible version of the picture theory of meaning. The basis for this argument was a distinction between 'semantic mood' and 'sentence-radical'.29 The common component in sentences bearing different forces, i.e. what Hare had called 'the phrastic', Stenius relabelled 'the sentence-radical' to emphasize that this entity, like a radical in chemistry (e.g. the hydroxyl group), cannot occur on its own, i.e. cannot be used to 'make a move in the language-game'. He represented the sentence-radical by an indirect statement (a 'that'-clause). This expresses the entire descriptive content of a type-sentence; it shows the state of affairs which the sentence describes. In paraphrasing a sentence, the sentence-radical must be completed by affixing to it a 'functional compo­ nent'. This signals the 'semantic mood' of the sentence to be paraphrased. It makes explicit in paraphrase what communicative act the original sentence performs by specifying the role played by the sentence-radical in presenting a state of affairs. Stenius argued that the picture theory of the Tractatus could stand intact as a theory of the meaning of the sentence­ radical. But it must be supplemented by a theory of semantic moods to account for the different uses of sentences in various 'language-games'. His apparatus of six mood-operators was intended to give a precise form to this programmatic claim allegedly presented in the Philosophical Investigations. 30 A similar ambition is evident in many modern speech-act theories of meaning. They attempt to embroider upon Austin's distinction of sense and illocutionary force. By characterizing both the sense and the illocution-

ary force potential of every sentence, they hope to give an exhaustive account of the workings of a natural language. Modern suspicion about abstract objects has tended to motivate a purge of Platonic entities such as senses or propositional-contents. Instead, it is frequently claimed that there is a basic type of 'propositional act' which is an ingredient of every complex type of speech-act, and that a particular propositional act may be a common component of different speech-acts of asserting, commanding, questioning, etc.3 l Understanding an utterance consists in grasping what propositional act the speaker performed as well as what illocutionary act he effected (making an assertion, issuing an order, etc.). It is assumed that such understanding must be derived at least in part from features of the type-sentence uttered by the speaker. Hence it is assumed that

31 Cf. J. Searle, Speech Acts (Cambridge University Press, Cambridge, 1 969), pp. 22-6, and J. Lyons, Semantics, vol. 2, (Cambridge University Press, Cambridge, 1977), pp. 735ff.

2. E. Stenius, Wittgenstein's Tractatus (Blackwell, Oxford, 1 960), ch. ix. 30 This is a grotesque misinterpretation of the Philosophical Investigations; d. G. P. Baker and P. M. S. Hacker, Wittgenstein: Understanding and Meaning (Blackwell, Oxford, 1980), pp. 1 1 0f£.

33 Kenny ( Will, Freedom and Power, pp. 39ff.) suggests that connections between utter­ ances which are disguised beneath ordinary expressions can be highlighted with such a notation, e.g. internal relations among the expressions of mental states. 34 Searle, Speech Acts, p. 3 1 .

65

[w Je can distinguish two (not necessarily separate) elements in the syntactical structure of a sentence, which we might call the propositional indicator and the illocutionary force indicator. The illocutionary force indicator shows how a prop­ osition is to be taken, or, to put it another way, what illocutionary force the utterance is to have; that is, what illocutionary act the speaker is performing in the utterance of the sentence. 32 A complete theory of meaning would give a perspicuous representation of every type-sentence by indicating both what propositional act and what illocutionary act would typically be effected by its utterance. This en­ terprise is thought to be furthered33 by paraphrasing sentences into a formal notation. We can represent these distinctions in the following symbolism. The general form of (very many kinds of) illocutionary acts is F(p), where the variable 'F' takes illocutionary force indicating devices as values and 'p' takes expressions for prop­ ositions. We can then symbolize different kinds of illocutionary acts in the forms, e.g. I- (p) for assertions Pr(p) for promises

32 Searle, Speech Acts, p. 30.

!(p) for requests w(p) for warnings ? (p) for yes-no questions. 34

66

Sense and Force

The propositional act common to different speech acts can itself be analysed in terms of referring and predicating. But, of course, these are just the matters which truth-conditional semantics claims to account for. Hence, if that claim were acknowledged, the clarification of the truth­ conditions of the 'proposition' expressed by a sentence would be the major part of the business of a general theory of meaning for a language. The exact contours of the theory of illocutionary forces and force-indicators would remain to be worked out. These developments in philosophy have caught the attention of theoreti­ cal linguists, and a distinction between sense and force (under various labels) is commonly thought to be of decisive importance for the empirical study of languages. One reason for this transfer of ideas is that philosophers have tended to trespass into the area that linguists had staked out as their distinctive territory. Speech-act theorists speak of illocutionary-force-indicating devices in sentences of a natural language which enable a competent speaker of a language to recognize what il­ locutionary act is performed by the utterance ora sentence. But empirical semantics addresses itself to the essentially psychological question of how speakers and hearers 'figure out' the utterance-meaning of sentences used in a given context.35 Hence, to a theoretical linguist, there is no possibility of erecting a purely philosophical speech-act theory of meaning. Either what speech-act theorists have sketched is an acceptable general framework for an empirical semantic theory or it is rubbish. The first alternative is inherently more attractive. Theoretical linguists secure the widest scope for further research if they maintain that speakers calculate the meanings of utterances by combining contextual information with their knowledge of grammatical principles and word-meanings. Appa­ rently speakers must recognize a given utterance as an instance of a type-sentence the sense and illocutionary force potential of which are alike derivable from knowledge of grammar and word-meanings, and then they must 'contextualize' this calculation to yield knowledge of what was said and the illocutionary act performed by the utterance. (The operation of this mechanism is, to be sure, very fast!) Philosophers, pontificating from their armchairs, may have outlined the proper research programme for linguistics, but any further progress requires gathering empirical knowledge about grammatical depth-structures which only the linguist is qualified to provide. The other motive for linguists' taking an interest in philosophical discus35 Katz, Propositional Structure and Illocutionary Force, p. xiii.

The Evolution of the Species

67

sions of sense and force arose from the growth of transformational­ generative grammar following in the footsteps of Chomsky. The distinction between la langue and la parole is safeguarded by drawing a sharp line between semantics and pragmatics. Pragmatics is held to encompass the 'uses and effects of signs within the behaviour in which they occur', whereas semantics requires that one 'abstract from the user of the language and analyze only the expressions and their designata'.36 Conse­ quently the focus of linguistics on semantics (and syntax) had been con­ sidered, since Saussure, to demand the exclusion from attention of such matters as what entity (if any) is referred to by a particular token of a context-dependent expression or what speech-act is effected by the utter­ ance of a sentence on a particular occasion. Linguists had to take note of the syntactical differences between declarative, interrogative, and impera­ tive sentences, but they relegated the study of correlations between sentence-forms and speech-acts to the separate (and unesteemed) science of pragmatics. This way of demarcating pragmatics had the effect of licensing linguists to ignore most issues about the moods of verbs or the forms of sentences in building semantic theories. Chomsky'S work did not remove the inclination to insist on a sharp boundary between semantics and pragmatics; on the contrary, his distinc­ tion between competence and performance called for this form of defence. But his description of syntax in terms of transformations did suggest an awkwardness in the traditional way of drawing this boundary. Transform­ ations can be described which carry direct statements into indirect speech, or assertions into the antecedents of counterfactuals, etc. And Chomsky suggested that an adequate semantic theory would take the form of assigning meanings to simple expressions and then explaining how this meaning is passed through transformation-operations to the infinite variety of well-formed sentences of a natural language. But transformations are also available to carry assertions into sentence-questions, into Wh­ questions, and (often) into imperatives. Hence the general thesis that meaning is transmitted through grammatical transformations presup­ poses that non-assertoric sentences have meanings falling within the purview of semantics, and a semantic theory based on transformational­ generative grammar achieves a fully integrated and elegant shape if the syntactic differences between declarative, interrogative, and imperative sentences are mirrored in semantic differences. This obvious desideratum has been accomplished in a straightforward way by a draft on speech-act

3.

Quotations from Lyons, Semantics, vol. 1, p. 1 15.

Sense and Force

The Evolution of the Species

theories propounded by philosophers. To semantics is now allocated not only the analysis of whatever expresses the propositional content of a sentence, but also the study of force-indicating devices which are con­ ceived to fix the illocutionary-force potential of the sentence.37 A Choms­ kian theory of competence embraces the study of the knowledge that an ideal speaker-hearer has of the information about illocutionary-force potential which is present in the grammatical structure and lexical compo­ nents of a sentence. But it stops short of considering the actual use made of uttered sentences. This study is allocated to pragmatics. For it is held that the actual use or force of an utterance is inferred by speakers from combining knowledge of its illocutionary-force potential with knowledge about the circumstances of its utterance. Pragmatics is thus thought to deal with 'the various mechanisms speakers use to exploit the richness of context to produce utterances whose meaning in the context diverges predictably from the meaning of the sentences of which they are tokens'. 38 Pragmatics thus provides principles for bridging the gap opened up by theoretical linguists between understanding a type-sentence and under­ standing what is said by uttering a token of it. Distinguishing sense from force has become an entrenched strategy in modern theories of language. It has removed the major embarrassment of truth�conditional semantics. Indeed, it has converted an objection into a vindication of this framework for the analysis of meaning. In the course of so doing, however, it has transformed what apparently began as a series of familiar observations about moods of verbs and syntactical types of sen­ tences with familiar discourse functions, and about ordinary conventions of indirect speech and the uses of verbs of speech into something altogether unfamiliar. The ordinary mundane uses of language suddenly take on different aspects, and the mind whirls. Do all sentences really contain a phrastic in the form of a verbal noun? Are imperative sentences, interroga­ tives, optatives, really instruments for expressing true or false propositions (in addition to issuing orders, asking questions, expressing wishes)? Do we really, without noticing it, 'figure out' the meaning and force of an utter­ ance on the basis of our knowledge (only now revealed to us) of principles of depth-grammar concerning illocutionary-force potential and of principles of pragmatics? We might wonder whether senselforce theories are not appropriate objects for Dr Johnson's observation:

Wheresoe'er I turn my view All is strange, yet nothing new. Endless labour all along, Endless labour to be wrong.

68

37 W. P. Alston, Philosophy of Language (Prentice-Hall, Englewood Cliffs, NJ, 1 969), pp. 36ff. ,. Katz, Propositional Structure and IIIocutionary Force, p. 15.

3

69

Lateral radiation

The march of history has led to an apparent consensus. What seems essential is that we should have some division of sentential utterances into a determinate range of categories, according to the type of linguistic act effected by the utterance; that there should be some notion of the sense of a sentence, considered as an ingredient in its meaning and as capable of being shared by sentences belonging to different categories; that the notion of sense be such that, once we know both the category to which a sentence belongs and the sense which it carries, then we have an essential grasp of the significance of an utterance of the sentence; and that, for each category, it should be possible to give a uniform explanation of the linguistic act effected by uttering a sentence of that category, in terms of its sense, taken as given. I do not think that we have, at present, any conception of what a theory of meaning for a language would look like if it did not conform to this pattern. 39 A somewhat more attentive and critical review of the underlying history, however, reveals that this apparent consensus is a sham. Seeming agree­ ment masks fundamental disagreements as well as deep unclarities. Some hints about these problems have been dropped in the previous survey, but the issues must now be put under the spotlights. The first rift among theorists concerns the identification of the entities described as having senses and forces. One main branch characterizes sense and force as features of type-sentences. The force of a sentence is fixed by conventions about the significance of such force-indicators as the mood of the verb, syntactic structures, and the occurrence of explicit performatives such as 'I promise'; typically a type-sentence is thought to have a single invariant force, upon which the determination of the illocu­ tionary act performed in uttering it partly depends. Similarly, sense is held to be fixed by general conventions about the use of expressions in refer­ ences and predication, and what an utterance expresses is determined by the interaction of the sense of the uttered sentence with features of the

39 M. A. E. Dummett, 'Can Analytical Philosophy be Systematic, and Ought it to Be?', in Truth and Other Enigmas (Duckwotth, London, 1978), p. 450.

70

Sense and Force

context of utterance. On the other hand, some theorists openly apply a sense/force distinction not to type-sentences, but to particular utterances. Frege held this view. The sense of different tokens of a single type-sentence may differ. And his critical thesis that the indicative mood and declarative sentence-form is at best a fallible guide to assertoric force makes sense only if he restricted assertoric force to utterances of a sentence actually used to assert the thought expressed. Austin too explicitly ascribed illocutionary force to utterances, conceding the possibility that tokens of a type-sentence may differ in illocutionary force (e.g. an utterance of 'There is a bull in the field' might have the force of a warning or of a description). Others follow this tradition.40 What it makes sense to say about sense and about force depends on which of those two approaches a theorist adopts. A related conceptual disagreement is associated with the term 'force' (and its congeners). On one view, force is a feature of a type-sentence which may restrict the range of possible speech-acts that can be performed by utterances of this sentence and which partially determines what speech­ act is actually performed by a particular utterance. Here force is held to be related to the actual use of a sentence as potentiality to actuality. Force may thus be conceived to characterize the typical use of utterances of a sentence, although the circumstances of a particular utterance may cause the actual use to diverge from the typical one. Accordingly, a sentence­ question may be ascribed invariant interrogative force even though particular utterances of sentence-questions may be used (and understood) to perform other speech-acts (e.g. to make a request or, as a 'rhetorical question', to make an assertion). On another, antithetical, view force is a feature of an utterance which characterizes the actual use of a sentence on a particular occasion. Accordingly, there is no such thing as divergence between force and use. Hence an utterance correctly classed as a rhetorical question would have assertoric, not interrogative force, and 'Can you pass the butter?' has the force of a request not of a question. On this view, what are picked out as 'force-indicators' in sentences, e.g. moods of verbs and even explicit performatives, are at best fallible guides to the force of an utterance, for it is expressly conceded that declarative sentences or 'performative utterances' may be used to effect a variety of different speech-acts. Once again, what it makes sense to say about force depends on which explanation is given of this term of art. The diversity of the concepts of force which are in play among theorists 40 e.g. Alston, Philosophy of Language, p. 37, and D. Davidson, 'Moods and Perform­ ances', in Meaning and Use, ed. A. Margalit (Reidel, Dordrecht, 1979), p. 17.

The Evolution of the Species

71

is most conspicuous in respect of the treatment of the identification and differentiation of forces. Some maintain that there are but two, the as­ sertoric and imperative (the 'fiat mood'41), arguing that the whole gamut of different speech-acts can be reduced to these simple elements.42 Others wish to add another force or two to this primitive stock, because, for example, they refuse to assimilate questions to imperatives, or they con­ sider expressions of intention to differ fundamentally from both assertions and commands (in particular, such sentences as 'I'll go' or 'I'll do it' are not to be assimilated to self-addressed imperatives). Yet others identify more distinct forces or 'moods'. 43 Finally, some might argue that the attempt to reduce the diversity of speech-acts to the operation of a handful of 'moods' of sentences 'on analysis' or 'in depth-grammar' is misguided. Rather are there as many distinct forces as there are speech-act verbs which differ in meaning.44 In short, theorists of force are in radical disagreement, not merely about the facts of language and speech they purport to describe and explain, but about the very concepts they should use in so doing. An exactly parallel ambiguity besets the term 'sentence-radical' (and its cousins). On one view, a single sentence-radical is associated with each unambiguous type-sentence. On another, different sentence-radicals may 41 The term 'mood' is systematically misused in current debates, its nexus with the morphology of the verb being altogether abandoned. 42 Kenny, Will, Freedom and Power, pp. 38ff.; d. Hare 'Meaning and Speech Acts'. Kenny distinguishes the 'assertoric mood' from the 'fiat mood' by reference to the 'onus of match' between a sentence and the state of affairs it describes. If the state of affairs described does not obtain, do we fault the sentence or the world? If the former, then the sentence is in the assertoric mood, if the latter, it is in the mood of fiats. This test is more problematic than it appears to be. The dichotomy of 'faulting' the sentence or 'faulting' the world is neither clear nor evidently exhaustive. Its application to normative sentences is opaque, e.g. 'You ought to cp', 'You must cp', 'You have a duty to cp', 'You have a legal obligation to cp'. What state of affairs do these sentences describe? - the addressee's cping, or his being required to cp? In these cases, if the addressee does not cp, do we fault the sentence or the world? So, too, 'Promises ought to be kept', 'All men have a right to freedom of speech' present problems. But even in the simple imperative case, if a commander issues a rash order, or an infelicitous order (e.g. to shut a shut door) and the 'state of affairs it describes' does not obtain (because, e.g. the troops get killed trying to obey it, or the addressee, unsurprisingiy, cannot shut shut doors) do we fault the sentence or the world? We certainly fault the commander! So too in the declarative case, if a speaker asserts that the cper 1/1s, and there exists no I and a theory of meaning will both clarify how this derivation is possible, and what the formal nature of the processes in­ volved in understanding are. The full-blown explanatory enterprise lies at the heart of the new science of theoretical linguistics. For, in accord with a tradition inaugurated by Saussure, Chomsky and the majority of his followers conceive of linguis­ tics as a branch of psychology. The grammar which the linguist constructs is not merely a predictive device for hypothesizing the native speaker's 'linguistic intuitions', but it is also held to be something the 'speaker­ hearer' implicitly knows. ' [G]rammar is a system of rules and principles that determine the formal and semantic properties of sentences. The grammar is put to use, interacting with other mechanisms of the mind, in speaking and understanding language.' 22 As we have seen in the previous chapter, Chomsky's 'theory of understanding' is rich, elaborate and fantastical beyond anything to be found in the pages of philosophical theorists of meaning. Although, as we know, the child's knowledge of grammar is a matter of 'mental processes that are far beyond the level of actual or even potential consciousness', 23 nevertheless a scientific investi­ gation reveals that the child's mind is richly furnished with rules of grammar, e.g. 'the child's mind (specifically, its component LT(HL) ) 24

20 21

Dummett, 'What is a Theory of Meaning? (II)', p. 70, our italics. Dummett, 'What is a Theory of Meaning?' p. 1 12, our italics (ef. p. 109, and 'What is a Theory of Meaning? (II)', p. 69). 22 N. Chomsky, Reflections on Language (Fontana, London, 1 976), p. 28. 23 N. Chomsky, Aspects ofa Theory ofSyntax (MIT Press, Cambridge, Mass., 1965), p. 8. 24 This is the scientific way of writing 'Learning Theory of Human Languages'; this theory is, of course, known to the child (implicitly!).

326

The Generative Theory of Understanding

The Generative Theory of Understanding

contains the instruction: Construct a structure-dependent rule, ignoring all structure-independent rules' /5 and 10 and behold! the infant obeys this instruction and constructs as part of its grammar a structure-dependent rule for converting declaratives into interrogatives. The grammarian is at work trying to discover the system of rules, syntactic, semantic and prag­ matic, which we all allegedly know (tacitly), and constantly use when we speak, think or understand the speech of another person. For it is held that we 'derive' our understanding of a sentence by 'computing' its meaning:

the overall similarities are striking. A composite Galtonian picture of the enterprise may be unjust in detail to specific theorists, but it serves to illustrate the general nature of their endeavour. For our concern is not to question the details of these grand theories, but their fundamental pre­ suppositions, what they have in common and what is never called into question because it seems self-evident.

What a speaker does when he understands or produces an utterance must include at least the implicit analysing of its syntactic structure. It is this ability that a theory of syntax seeks to explicate . A semantic theory takes the solution to the general problem of production and understanding a step further. It seeks to account for the speaker's ability to assign interpretations to sentences on the basis of his knowledge of the meanings of their parts . . 26

The various programmes for full-blown theories of language have in common the idea that the units or elements upon which formation- and transformation-rules work are typically words or 'lexical items' and that our understanding of sentences is 'constructed' or 'derived' from our grasp of word-meanings. Many myths lie hidden behind this seemingly in­ nocuous supposition, and we shall expose a small selection. We shall disregard phonological and morphological considerations as being of less obvious philosophical concern than semantic ones. We shall also pass over substantial difficulties27 that arise for the rule-obsessed generative gram­ marian with semi-productive formations that are irregular and not sharply circumscribed (e.g. adding '-ness' to an adjective, which applies happily to 'white', 'black', 'blue', but unhappily to, say, 'purple' and not at all to 'magenta' or 'eau-de-nil' ). We shall begin with words, taking the term in a down to earth manner. It will be helpful to start with an outline of the strategy pursued by one group of linguists, although it should be noticed that our criticisms raise points which have to be answered by any enter­ prise that purports to 'derive' or 'construct' sentence-meanings from the meanings of constituent words and their form of combination. Words are 'the atoms of the syntactic system'/s for the fundamental idea is 'that the process by which a speaker interprets each of the infinitely many sentences is a compositional process in which the meaning of any syntactically compound constituent of a sentence is obtained as a function of the meaning of the parts of the constituent.' 29 These syntactically elementary constituents of a sentence are, in the grammarians' jargon, 'assigned a semantic representation'. This is done by a 'dictionary' (not a book, but a 'theoretical construct' consisting of all the rules assigning meanings to words in the language). According to advocates of so-called

.

.

.

The full-blown enterprise incorporates the minimalist one in a qualified sense. It would be wrong to suggest that any theorist pursuing a theory of understanding which postulates tacit knowledge of a theory of meaning must also claim that the theory of meaning itself constitutes a 'theoretical representation of a practical ability' or a 'recovery of the structure of a very complicated ability'. He might deny the very intelligibility of this claim. But a theory of meaning which will deliver the possibility of specifying 'in a systematic manner' the meaning of any sentence of the language is a common instrument in the achievement of the divergent objectives. For the time being we shall focus exclusively upon the full-blown enterprise. We do so for three reasons. First, it is this above all which purports to answer the question of how we can understand sentences which we have never heard before. Secondly, it is intellectually more ambitious, linking understanding with mental processes, and flaunting explanatory pretensions in psychology. Thirdly, it (and not the minimalist enterprise) is the primary point of contact between contemporary philo­ sophy of language and theoretical linguistics; hence a critical investigation may pay double dividends. We shall return to consider the minimalist position only at the end of our primary investigation. Even within the camp of theorists engaged in the full-blown explanatory enterprise, the differences between linguists and philosophers are not trivial. Even the disagreements within each group are not insignificant. Yet 25 Chomsky, Reflections on Language, pp. 32f. 26 Fodor and Katz, 'What's Wrong with Philosophy of Language?', p. 282.

3

327

The elements

27 P. H. Matthews, Generative Grammar and Linguistic Competence (Allen & Unwin, London, 1 979), pp. 27ff. 2. J. J. Katz, The Philosophy of Language (Harper & Row, New York, 1966), p. 153. 2. Ibid., p. 152.

The Generative Theory of Understanding

The Generative Theory of Understanding

'componential analysis' the meaning of a lexical item 'is not an un­ differentiated whole. Rather, it is analysable into atomic conceptual elements.' 30 Meanings of words 'are not indivisible entities but, rather, are composed of concepts in certain relations to one another, the job of the dictionary [being] to represent the conceptual structure of the meanings of words'. 3 1 The semanticist's analysis of the lexical items is explicitly com­ pared to the chemist's analysis of chemical compounds, and the former's diagrammatic representations of his analyses are held to be precisely parallel to the latter's chemical notation. This naive composition and the accompanying bizarre ontology do little to encourage confidence in the sanity of the proceedings, but we shall disregard this dangerous and objectionable style of thought. The fundamental idea is independent of the ontology, and it is the former which we aim to evaluate. The analysis of a word's meaning represents it by 'semantic markers'. Thus to the word 'bachelor' there will correspond four different entries to budget for semantic ambiguity which 'has its source inthe homonymy of words': 32

appears to be a general presumption that the non-logical vocabulary of a language can be divided into two grand classes, viz. definables and indefin­ abIes. The definable terms are to be defined or analysed into their charac­ teristic marks (or Merkmale) or by some other formal definition of a more complex kind, as in definitions of an ancestral relation. In all such cases, since the meaning of a word is held to consist of its contribution to determining the truth-conditions of any sentence in which it may occur, its meaning is held to be specifiable by stipulating conditions necessary and sufficient for its application. However, when it comes to the meaning of 'indefinables' the theorists are remarkably silent. Some assume that homo­ phonic T-sentences are explanations of meaning that link language to reality. Others perhaps assume that an ostensive definition is the primary means of explaining indefinables, and thus it is this device which links language to reality. And some apparently presume that indefinables can­ not be explained at all, but rather a grasp of their meaning consists in an ineffable recognitional capacity. It is perhaps no coincidence that the grammarian's Boolean analyses are tailored to the operations of a digital computer, and that the philosopher's story is so well adapted to the forms of the predicate calculus. Be that as it may, these regimented programmes must face two kinds of difficulties which have been little discussed, if at all, by these theorists. (i) A huge variety of words of many different types do not lend them­ selves to explanation along the lines required by these theorists. A few examples of such kinds of expression are as follows:

328

(i) (Physical Object), (Living), (Human), (Male), (Adult), (Never Married). (ii) (Physical Object), (Living), (Human), (Young), (Knight), (Serving under the standard of another). (iii) (Physical Object), (Living), (Human), (Having the academic degree for the completion of the first four years of college). (iv) (Physical Object), (Living), (Animal), (Male), (Seal), (Without a mate at breeding time). The semantic markers are not words or even linguistic expressions, although the grammarian 'represents' them thus. Rather, 'they are to be regarded as constructs of a linguistic theory, just as terms such as " force" are regarded as labels for constructs in natural science'. 33 Philosophers often subscribe to a more schematic method of assignment of meaning to constituent non-logical expressions in a language. There 30 J. J. Katz and P. M. Postal, An Integrated Theory ofLinguistic Descriptions (MIT Press, Cambridge, Mass., 1964), p. 14. 3 1 Katz, The Philosophy ofLanguage, p. 154. 31 Ibid., p. 153. 33 Ibid., p. 156. Numerous objections to this way of playing the game may cross the mind of any reader. Items (ii)- (iv) are not 'analyses' of the meaning of 'bachelor' at all, but of 'bachelor of art', 'bachelor-knight', and 'bachelor-seal'. Otherwise these latter three expres­ sions would be, in context, pleonastic, which they are not; they would be necessary only for disambiguation, but they are not. We shall disregard such qualms about detail in order to focus on more general objections.

329

(a) Summa genera such as 'substance', 'property', 'relation', and 'concept' are obviously problematic, not being readily analysable by Boolean semantic markers, nor by enumeration of characteristic marks or other formal definition. (b) Family-resemblance terms whose extension is not united under a unitary concept in virtue of shared characteristics stand equally in need of different treatment. In Wittgenstein's view numerous terms are of this kind, including such philosophically important ones as 'proposition', 'language', 'number', and 'understand'. (c) Psychological 'indefinables' must be problematic. For the only way (apart from behaviourism) that seems available to the theorist is to fall back on a form of private ostensive definition. Some will doubt­ less do so, but they must first demonstrate the intelligibility and coherence of this procedure as a method of explaining the meanings of terms of a public language.

330

The Generative Theory of Understanding

(d) Indefinables whose meaning is explained by reference to public samples, whether canonical, as in the case of the standard metre (or yard, etc.) or standard, as in the case of shades of colours, materials at a drapers, etc. (where books of samples are used). Whether these theories make room for such terms is debatable. The grammarian's mental lexicon clearly does not, since the mind can contain no standard metre rods or samples of colour or cloth. If the philosopher's theory allows room for such terms, he owes his audience an account of how these forms of explanation fit into an axiomatic calculus. Many other kinds of neglected expressions are no less problematic. Expletives, greetings, vocatives, prepositions, modal auxiliaries, and hosts of other parts of speech whose uses are familiar and readily explicable, are neither explained nor obviously explicable in the stipulated forms. (ii) A quite different objection would challenge the supposition that there are clear principles of individuation for word-meanings. It is sup­ posed that meanings or 'interpretations' are 'assigned to' or 'paired with' words, and that every significant use of a word in a sentence involves that word being paired with one meaning or another. But is this not a piece of mythology that stems from viewing language as a formal calculus which is given an interpretation? For quite apart from objections to the bi-planarity of the conception, there are no absolute, definite, principles for counting how many meanings a word has. Take the verb 'to play' and observe but a small segment of its ordinary, richly diverse use: He plays cricket. He plays the ball straight down the wicket. He plays Mercutio. He plays the violin. He plays the second violin sonata. He plays Mozart. He plays the fool. He plays fair (or foul). He plays himself in. He is played out. He is playing up. He plays them off against each other. He plays with women. He is playing into her hands. He plays his cards well. He plays on words.

f 4:

The Generative Theory of Understanding

331

England will play its part. The light plays upon the water. What meaning is 'associated' or 'paired' with the verb 'to play'? Or is it ambiguous? If so, how many meanings does it have? It might be thought that the example is exceptional, perhaps because 'play up', 'play down', etc. are to be treated as distinct and semantically indivisible (like prepositions separated from verbs with separable prefixes in German) or because 'to play' is a family-resemblance term. But the phenomenon is ubiquitous. The word 'air' is surely not a family­ resemblance term. But are there any agreed non-arbitrary ways of count­ ing how many meanings it has? For one may air one's blankets before the fire, but one's opinions before the company; however, one would be ill-advised to air the room in the evening, for the night air is damp. One may rejoice to return to one's native air, delight to take the air in the afternoon, find, once out in the open air, that the air is sweet in the fields. Since there is a light air from the south, one may go sailing, and crowding on canvas, sail with gentle airs for a few miles. But the gentle airs with which one sails are not the airs one puts on. Those who do not put on airs may have a warm air, although their chatter may be a lot of hot air. The air we breathe may be cool, and if we are on Everest, thin. Yet if a ghost melts away into thin air, it may be at sea-level for all that, not thousands of feet up in mid-air. Although we may appear to be building castles in the air, since all this has an air of paradox, the points we are making have been in the air for some time, though not often heard on the air. Obviously, an actual dictionary must use some ad hoc principles for determining the number of entries for an expression. But no lexicographer would claim special status for his principles, and different dictionaries will employ different principles. The generative grammarian's hypothesized mental lexicon, however, can hardly allow such flexible, pragmatic, p rinciples in an enterprise that claims explanatory force. For while an ordinary lexicographer merely attempts to explain the uses of a word, typically by invoking hosts of illustrative examples, the language theorist attempts to explain both understanding and derivation of sentence­ meaning from the composition of lexical items whose meanings must be laid down in advance of composition. 34 The mind, be it noted, is here conceived as a very powerful computer! 34 It might be argued that since the hypothesis is underdetermined by the data, the theorist must simply choose the best hypothesis. But it is altogether absurd to suppose that 'the mind' has a set of principles for the individuation of numbers of meanings words have, as opposed to the lexicographer deciding on ad hoc principles relative to his purposes.

The Generative Theory of Understanding

The Generative Theory of Understanding

The two general kinds of difficulty suggest that the task of 'discovering' the entries in the theoretical lexicon, or of stipulating the axioms of the theory of meaning, may prove harder than is commonly envisaged. But this may simply spur the theorist on to yet greater ingenuity and more strenuous efforts in his theory-constructing enterprise'. He might try to budget for family-resemblance terms by analysis into differentially weighted disjunctions of characteristic marks, the satisfaction of some weighted majority of which would license application of the expression. He might devise a hard "and fast criterion for determining numbers of dictionary entries for a given expression. He might postulate 'neural representations' of samples, on the argument that only the further development of his theory will show whether his postulate makes sense. All this, however, is of no avail. For the deep objection is not that the task is Herculean, but it is Sisyphean, because altogether misconceived. For even if the theorist can come up with appropriately complex explana­ tions, what would their normative sta1;us be? We explain family­ resemblance words by a string of examples plus a similarity rider, not by complex formulae involving weighted averages of semantic values, what­ ever they might be. We typically explain 'indefinables', such as colour predicates or predicates of perceptual qualities in general, by ostensive definitions by reference to paradigmatic samples. We typically explain prepositions by contextual paraphrase. We explain some expressions by contrastive paraphrase; others by exemplification. These are not indirect gestures towards the 'real' meaning of such terms, but genuine and correct explanations in their own right.35 The language theorist will, of course, acknowledge that we do not, in our ordinary explanatory practices, deliver explanations of word-meaning in the regimented form of axioms or semantic markers specifying necessary and sufficient conditions. But this, in his view, merely shows that an ordinary speaker does not have explicit knowledge of a theory of meaning which the linguist aims to reconstruct. The correct theoretical explanations will have the stipulated forms, for only thus can language be represented as a calculus which delivers the meanings of sentences as consequences of axioms. However, these axioms or dictionary entries are supposed to represent a speaker's tacit knowledge. For, as noted, the speaker is said to utilize his knowledge of the ' dictionary' to obtain the meanings of compound expres-

sions. The theorist presents the axioms of his theory as meaning-rules,26 or as semantic markers. Each lexical item has its meaning itemized by a unique rule (or, in the case of ambiguity, each meaning of a homophone is uniquely stipulated). But now notice a further discrepancy with our ordinary conception of explanation. We not only give explanations in forms not acknowledged as legitimate for the calculus of language, but we recognize as legitimate different explanations of one and the same un­ ambiguous expression. The term 'circle' can be explained as 'the shape of a penny piece'; or it can be explained ostensively by pointing at a circular object and saying 'A circle is this i shape'; or it can be explained by exemplification: 'Look, this " (drawing in the sand) is a circle'. And a geometer might explain it as 'the locus of points equidistant from a given point'. 3? Wholly disregarding other uses of 'circle' (e.g. 'The Vienna Circle', 'the endless circle of parties', 'in such social circles', etc.), which of these explanations of the meaning of 'circle' is the native speaker alleged to know? If a person is asked what 'circle' means, and points at a plate, saying 'The shape of that i is a circle', does this reveal his knowledge of the 'dictionary entry' or 'axiom' for 'circle' or not? Or is it that he uses some other unique and regimented form of definition, which he implicitly knows, and on which basis he gives such 'unsatisfactory' overt explana­ tions? If the former, then the tales of axioms and lexicons are redundant. For what the speaker knows when he knows the meaning of a given word is just what he explains when he explains its meaning (whether by osten­ sion, formal definition, exemplification, examples, paraphrase, con­ strastive paraphrase, or whatever) and what he acknowledges as a correct explanation when he encounters one. But if the theorist insists, as he doubtless will, that the speaker's typical explanations are incorrect, at best only ersatz versions of the correct theoretical axioms or semantic markers, then he wilfully opens a vast gulf between our ordinary explanatory practices and the 'explanations' of the theory. The theorist will indeed do so, in the name of scientific progress. Theory-construction in linguistics is no more constrained, in his view, by our common-or-garden explanations of meaning than theory-construction in physics is constrained by common-or-garden explanations of motion,

332

3S For more detailed elaboration of these points, see G. P. Baker and P. M. S. Hacker, Wittgenstein: Understanding and Meaning (Blackwell, Oxford, 1980), pp. 69ff.

333

36 The theorist might deny that his axioms are rules. But by that very token he must, self-defeatingly, deny that they 'give', or explain, the meanings of words. For explanations of word-meanings are normative, they specify the correct way of using an expression, deviation from which, in a given context, is a mistake. 37 It might indeed be argued that, far from the geometer's definition being the 'correct' one, it defines a somewhat different concept of circle from the common-or-garden one.

The Generative Theory of Understanding

334

force or acceleration. The correct form and uniqueness of a semantic axiom or a conglomeration of semantic markets are dictated by its role in a powerful explanatory theory. The actual explanations given by a speaker are mere inductive evidence for his tacit knowledge of the 'dictionary entry'. However, borrowing the jargon of advanced science does not make speculative linguistics into superphysics. Inductive evidence presupposes non-inductive identifications and subsequent correlations. If the linguist invokes a putative . analogy with 'theoretical entities' such as electrons, whose existence is allegedly 'postulated' by physicists, we should remind him that one can weigh them, measure their velocity or electric charge, and manufacture or annihilate them. But the analogy is not merely weak, it is altogether misplaced. For the theorists' 'dictionary entries' or 'axioms' are in effect hidden rules conceived to operate at a distance. They are 'tacitly known' or 'cognized'. Even though buried at inaccessible depths of uncon­ sciousness, they are used whenever we speak or understand what is spoken. But this is to fall victim to a mythology of rules. Rules cannot 'act at a distance'; there is no such thing as a rule determining behaviour unbeknownst to anyone who might cite it, consult it, and invoke it in explanation, criticism or justification. To think that rules await discovery by linguistic theorists is to confuse the appropriate forms of explanation of normative phenomena with forms of explanation appropriate only to the physical sciences. If explanations of word-meaning are rules for the correct use of words, at any rate these explanations cannot, on pain of in­ coherence, be represented as objects, awaiting discovery, which have no overt use in public linguistic transactions. Nor can they be unintelligible to those who are alleged to use them (as would be, e.g., Frege's explanations of the meanings of number words), for then they could not use them. A rule is not a predictive device, and whether a speaker does or does not guide his linguistic behaviour by reference to a given rule is not decided by a prediction that he will behave in such-and-such a manner. Consequently, to the extent that the theorist tries to meet objections of one of the two general kinds adumbrated above by constructing more and more elaborate hypotheses about wonderful and hitherto unknown rules, his ingenuity is misplaced and his efforts misguided.

4

, ;,�

The Generative Theory of Understanding

displayed as derived from the meanings of the words combined. But not any combination of words has a meaning. And given that many, perhaps most, words have multiple entries, theorists are inclined to seek out general principles whereby we select an appropriate entry for a given combination of words. This selection of senses and exclusion of others is reconstructed in the semantic component by the device . . . [of a] . . . selection restriction. Selection restrictions express necessary and sufficient conditions for the readings in which they occur to combine with other readings to form derived readings.3•

These selection restrictions can be formulated as Boolean functions of semantic markers, e.g. the restriction on 'honest' for its reading 'charac­ teristically unwilling to appropriate for himself what rightfully belongs to another, avoids lies, deception, etc. ' is the Boolean function « (Human) & (Infant) where the bar over the semantic marker indicates exclusion in the reading in question; for its archaic reading 'chaste', the selection restric­ tion in written « (Human) & (Female) . Projection rules are then held to stipulate the manner of amalgamating readings of individual words to give readings for phrases and sentences. The projection rules of the semantic component for a language characterize the meaning of all syntactically well-formed constituents of two or more words on the basis of what the dictionary specifies about these words. Thus, these rules provide a reconstruction of the process by which a speaker utilizes his knowledge of the dictionary to obtain the meanings of any syntactically compound constituent.3•

This 'theory' not only explains the 'process' of understanding, it is also allegedly rich in predictive power: We can now show how a semantic component of English can predict . . . facts about the semantic anomaly of sentences such as 'The grain of sand is good' . . . etc. and the non-anomalousness of sentences such as 'The razor blade is good' . . . etc. Sentences of the former type have subject nouns whose reading contains no occurrence of an evaluation semantic marker and thus, in their semantic inter­ pretation, these sentences receive no readings because the readings of their subjects do not satisfy the selection restriction in the reading for 'good'.4"

Combinations of elements: phrases and sentences

The abstract lexicon or the axioms of the theory of meaning represent the meanings of words. The meanings of combinations of words must be

335

38 Katz, The Philosophy of Language, pp. 159f. 39 Ibid., pp. 1 61£. 40 Ibid., p. 297.

336

The Generative Theory of Understanding

Variations upon this form of 'componential analysis' were common among linguists in the 1960s and 1 970s. Other types of analyses, however, proliferated. And controversy raged over whether transformations do or do not change meaning, and whether the 'model' of linguistic description should be 'syntactically based' or 'semantically based'. None of this need concern us, since our qualms arise at a more elementary level than these internecine feuds. It is clear enough that if one is embarked upon the construction of any theory of meaning which will enable the derivation of the meaning of any sentence from the meanings of its constituents and mode of combination, one must take steps to exclude nonsense or semantic anomaly. But it is unclear what falls under this heading, and theorists disagree radically among themselves about how to draw the bounds of sense. What counts as nonsense can be variously characterized. One might call a type-sentence nonsensical only if it has no intelligible use in any conceivable circum­ stance; or only if the established practice of explaining meanings gives it no standard use. Alternatively one might call an utterance nonsensical iiit is unintelligible to competent speakers, even if other utterances of the same type-sentence in different circumstances would be intelligible. Likewise, the criteria for judging an expression to be nonsensical can be differently stipulated or described. One might require that competent speakers assent to the metalinguistic statement ' "p " is nonsense'; or that utterances of the problematic sentence standardly evoke responses of incomprehension; or that a particular utterance is unintelligible to a competent speaker. Theorists of meaning seem oblivious to these differences. They evidently consider that the sense/nonsense distinction marks circumstance-invariant features of type-sentences, and they show no awareness that different standards, appropriate to different intelligible purposes, might yield discrepant results. They hanker after general principles which will draw a once-for-all dichotomy between type-sentences, and they assume that a sentence which makes sense will invariably be intelligible to a competent speaker. To the extent that compositional theories of meaning are committed to these presuppositions, they are rooted in nonsense and confusion. There are no tolerably general principles which lay down necessary and sufficient conditions for combinations of words to make sense. It is notorious that attempts to construct general theories of categories tend towards the absurd conclusion that there are as many categories as there are words,41 41 Cf. W. and M. Kneale, The Development of Logic (Clarendon Press, Oxford, 1962), p. 6 7 1 .

r , ,

I " w ,�

1;

�.

f

The Generative Theory of Understanding

337

and in any case competent speakers do not invoke any such rules in j ustifying their discrimination of sense from nonsense. The idea that a sentence invariably makes sense or invariably fails to make sense is chimerical, and defending this dogma by categorizing intuitively ill­ formed sentences as necessarily false leads from one absurdity into another. These are serious difficulties for compositionalism. For it would be preposterous to retreat to the position that any combination of individually significant words makes sense. The context-dependence of sense and nonsense was argued in Chapter 6. A pair of further illustrations will drive the point home. To understand the quip 'the worst passions of the human mind are called into action by the pulverists and the lumpists', one needs to know that Sydney Smith is recounting an argument between those who favoured sweetening their tea with powdered sugar and those who favoured lump sugar.42 Out of context, the following might pass for nonsense: Victory has got a half-Nelson on Liberty from behind. Liberty is giving away about half a ton, and also carrying weight in the shape of a dying President and a brace of cherubs. (One of the cherubs is doing a cartwheel on the dying President's head, while the other, scarcely less considerate, attempts to pull his trousers off . . Y

But in its context this is immediately understood to be a wickedly witty description of a nineteenth century memorial statue (to a Brazilian presi­ dent in Rio de Janeiro). The circumstances surrounding an utterance often make the crucial difference between grasping what is meant and not being able to conceive what on earth is conveyed. A stronger objection to compositionalism must also be sustained. Such a theory of meaning must presumably classify as nonsense any combination of words which would standardly be unintelligible. Such a combination as 'dead rainbow' would assuredly be excluded by selection restrictions (or other combinatorial rules) . But suppose a little boy sees diffraction rings on a film of oil floating on a puddle after a rainstorm, and exclaims 'Here is a dead rainbow'. Is his remark to be judged unintelligible because the phrase is nonsensical ? Or should the theorist conclude that contrary to our prior convictions, the phrase really made good sense all along? Both horns of this dilemma are unattractive. But it is only forced upon us by the absurd 42 Hesketh Pearson, The Smith ofSmiths (Folio Society, London, 1 977) p. 98. 43 Peter Fleming, Brazilian Adventure, quoted in J. J. Norwich, Christmas Crackers (Allen �ane, London, 1 980), p. 76.

The Generative Theory of Understanding

The Generative Theory of Understanding

presupposition that semantic rules uniformly determine the legitimacy of combinations of words in advance of their actual production and thereby fix their meaning in combination with 'projection rules'. This conception, while purporting above all to explain the nature and possibility of linguistic creativity, in fact grossly distorts and misrepresents it. For parallel to the empiricist'S picture of the Kaleidoscopic Mind, so too the grammarian's tale reduces creativity in the use of language to the realization of a range of combinatorial possibilities which are pre­ determined and delimited in advance by a set of tacitly known rules. This not only involves a misguided mythology of rules, it debases creative use of language by arguing that it is a mechanical application of a calculus of meaning-rules, which, as it were, contains in advance the possibility of that novel use. Linguistic creativity, whether by poet, scientist or child, lies in exploiting combinations of expressions not readily foreseen or laid down as correct by rules of the language. (Could the rules foresee the possibilities to which the speakers were blind ?) It is rooted in our genuinely creative capacity for seeing analogies, moulding language to new contexts, and imposing new patterns upon the phenomena we experience. If a compositional theory grants sense to any combination of words which might conceivably be used to say something intelligible, much of what passes for nonsense must be reckoned among expressions which really make sense, and it will be difficult definitely to identify any outright nonsense at all. If, on the other hand, nonsensical expressions are admitted sometimes to make intelligible statements, the theory must abandon the thesis that what a speaker says must be calculated from the sense of the sentence uttered. Compositional theories of meaning are stranded in no man's land, caught in a deadly crossfire. Their only apparent refuge is to embrace the less unpalatable alterna­ tive. They may advocate conceptual anarchy, denying the legitimacy of any discrimination of sense from nonsense. Or, more plausibly, they may argue that there are no selection restrictions proper to semantics; only syntactically ill-formed expressions are nonsensical, while combinations commonly excluded as nonsense are rather necessarily false. According to this view, such sentences as 'Green ideas sleep furiously', 'This geranium is honest', or 'Here is a dead rainbow' are anomalous only in the same way as are 'This mat is both round and square' or 'Two is prime and two is not prime'. This position lacks any sound supporting argument. One attempted defence is patently circular, since it falls back on compositionalism itself. Confronted by the phrase 'honest geranium', one linguist reasons:

after all, I know what properties an entity must possess in order to be properly described by such a phrase - namely that it must be a flower of the genus pelargonium which is fair and upright in speech and act. Now it might be objected that no one has ever encountered such an entity in the real world; but that is a pragmatic fact [sic!] about the way things happen to be in the world right now.44

338

339

But what justifies the claim that he knows what it would be for a geranium to be honest, apart from the fact that he knows what 'honest' and 'geranium' mean in other contexts ? Could he even imagine encountering an honest geranium ? At this point recourse to a second argument may seem attrac­ tive. The theorist may invoke fairy tales, in which inanimate objects talk, feel pain, fall in love, where geraniums may tell lies and ideas might be green. Surely, he may say, we can imagine these episodes and we do understand these stories. They may be false, but they cannot be nonsense, for then we should not understand them. Invoking imagination, however, is misguided, since part of the criterion for whether one can imagine such-and-such is precisely that specification of the object of imagination makes sense. Our powers of imagination give us no independent handle upon what makes sense, and the frontiers of the logically impossible, the conceptually absurd, and the nonsensical cannot be pushed back by daily exercises to stretch one's powers of imagination. To invoke our under­ standing of fairy tales begs the question in so far as clarification of what counts as so understanding fairy tales is problematic. One might rather argue45 that such applications of concepts are essentially secondary, as is evident if we try to envisage the application of 'honest' only to flowers. Neither of the two arguments to show that what passes for nonsense should be reclassified as 'necessarily false' is compelling. Composition­ alism backs itself into a dilemma about distinguishing sense from non­ sense. The only escape from the impasse is to think out afresh the relation between meaning and understanding.

5

Understanding and tacit knowledge

Hitherto our discussion has focused upon difficulties which must be confronted in constructing any theory of meaning which will display sentence-meanings as derived from the meanings of constituent words and 44 A. Radford, 'The Origins of Katzian Semantics', unpublished preface to a collection of articles entitled 'Katzian Semantics', in the Taylorian Library, Oxford University, p. 29. 4S Wittgenstein, Philosophical Investigations, § 282.

The Generative Theory of Understanding

The Generative Theory of Understanding

their mode of combination. Though we concentrated on linguists' com­ positional theories, many of our objections constitute difficulties that have to be surmounted by non-compositional theories too, e.g. theories that conceive of word-meanings as contributions to determination of truth­ conditions. It may well be that most, perhaps all, of the difficulties we have raised will be viewed as challenges to be overcome, as a spur to greater ingenuity and theoretical inventiveness. So to respond only makes sense to the extent that the full-blown explanatory generative theory of under­ standing ( 1 ) can coherently make use of the envisaged theory of meaning in giving an explanation of understanding, and (2) that it is intelligible to explain, in the requisite sense, how we understand sentences. Now we shall examine the first issue. The construction of a theory of meaning is held to involve a theory of understanding in so far as it is supposed that a speaker of a language has tacit knowledge of the theory of meaning. The theory of meaning is meant to replicate the theory tacitly known or 'cognized' by a speaker. Hence understanding is explained as involving an analysis of sentence­ constituents, a coordination of their contributions to the meaning of the sentence in which they occur, an analysis of the structure of the sentence (surface and 'logical' or 'deep' structure) and thence the derivation of the meaning of the sentence. This notion of tacit knowledge46 has never been properly explained. Indeed, in some cases, e.g. Evans's (above, p. 295), it is virtually obvious that there could not possibly be any good grounds for talking here of knowledge at all. It is true that the notion has been

variously employed throughout the history of philosophy. Plato's sugges­ tion in the Meno that the slave was recollecting a priori truths of geometry inasmuch as he had not learnt them might be thought to exemplify one form of allegedly tacit knowledge. The Leibnizian account of 'virtually innate knowledge' is perhaps another, slightly different, variety. Few, apart from theoretical linguists, would wish to revive these long dis­ credited conceptions. But something that might be termed 'tacit know­ ledge' can be of a quite different nature. For example, Ryle's distinction between knowing how and knowing that might be thought to intimate tacit knowledge of whatever practical principles are, in some sense, fol­ lowed in exercise of an ability. Quite differently, a psychoanalyst's patient, who, under analysis, comes to realize that his motives were related to a childhood trauma might also be said to possess tacit, but not explicit, knowledge of his motives. Again, a pianist who can play and recognize Bach trills and Chopin trills might be said to have tacit knowledge that the former begin above the printed note and the latter on the printed note. For although when asked, he does not give this answer, when told the answer, he immediately recognizes it as correct. These are all different kinds of case, some unquestionably misguided and all of them unilluminatingly described as cases of tacit as opposed to explicit knowledge. When the theorist of language attributes to every speaker of a language tacit knowledge of a theory he must, in the first instance, clarify what this means. First, what are the grounds for attributing to a speaker such tacit knowledge? That our speech can (perhaps) be mapped on to a complex calculus no more shows that we have been operating one than the mere possibility of mapping Zulu war dances on to chess shows that Zulu warriors are chess-players. The questions of what distinguishes tacit knowledge from explicit knowledge on the one hand, and of what distinguishes tacit knowledge from ignorance on the other, must be answered coherently. For it is far from obvious that theorists use of the term 'tacit knowledge' is intelligible. Like Locke's notion of tacit consent, tacit knowledge may be no more than a device for tacitly burying problems raised by the theory itself. If we are to ascribe tacit knowledge of a theory of meaning to a person there must be something that will reveal the difference between tacitly knowing the theory and total ignorance of it, something other than the mere fact of the speaker's correct discourse. Ifhis correct discourse is all that shows his tacit knowledge and incorrect discourse is all that shows ignorance, then the hypothesis that he can produce and understand sentences because of such tacit knowledge is both un testable and vacuous.

340

46 Even Dummett, when engaged upon the minimalist enterprise, invokes language appro­ priate only to the full-blown one: 'A theory of meaning of this kind is not intended as a psychological hypothesis. Its function is solely to present an analysis of the complex skill which constitutes mastery of a language, to display, in terms of what he may be said to know, just what it is that someone who possesses that mastery is able to do; it is not concerned to describe any inner psychological mechanisms which may account for his having those abilities. If a Martian could learn to speak a human language, or a robot be devised to behave in just the ways that are essential to a language-speaker, an implicit knowledge of the correct theory of meaning for the language could be attributed to the Martian or the robot with as much right as to a human speaker, even though their internal mechanisms were entirely different.' ( 'What is a Theory of Meaning? (II)', p. 70) If the minimalist task of the theory of meaning is merely to give a proper analysis of the speaker's ability, to display what it is that he can do, why is implicit knowledge of the theory of meaning which constitutes the theoretical representation of the practical ability atttibuted to the speaker? What is the criterion for possession of such implicit knowledge and what distinguishes it from total ignorance? Is there any more reason to attribute implicit know­ ledge of the grammar of our language to this robot than there is reason to attribute implicit knowledge of mathematics to a pocket calculator?

341

The Generative Theory of Understanding

The Generative Theory of Understanding

Secondly, grammarians and philosophers construct different theories of language. They are all designed, more or less, to match the linguistic 'output' of speakers (actual or ideal), to 'explain their linguistic intuitions' or to 'predict' their judgments of grammaticalness or meaningfulness. But, presumably, any sequence of behaviour can be mapped On to more than one 'rule' or system of 'rules'. So even if it made sense to attribute tacit knowledge of rules of a grammar to a speaker, how would we select between various possible rules ? Traditional grammarians, with peda­ gogical purposes, might select different rules in writing a grammar book according to ease of surveyability by learners. Logicians, with their interest in formalizing canons of inference, will concoct rules according to ease and perspicuity of representing valid inference (and, of course, there are many possibilities here). But such purpose-relative principles of choice in laying down rules are not open to our theorists, who are bent on 'discovering' rules. For they are concerned with revealing the hidden rules which a speaker knows, rules the knowledge of which explains the speaker's ability to speak and understand. Thirdly, the earlier arguments have demonstrated that the very notion of hidden rules exercising their sway at a distance, and discoverable only by scientific investigation, is unintelligible. At the phenomenological level of understanding it is obvious that we do not calculate, compute or derive the meaning of a sentence from the meanings of its constituents and their mode of combination, a fortiori not according to the rules of some hitherto unheard of grammar. Indeed, even after we have, with the aid of linguists and philosophers, rendered some (?) of this treasured knowledge explicit, we still do not engage in any calculations or computations, save in excep­ tional circumstances such as 'deciphering' a sentence involving multiple embedded relative clauses - and then we invoke more mundane grammars anyway (and perhaps the aid of paper and pencil). To this it will be replied that the whole point of introducing the notion of tacit knowledge was precisely because of these very phenomenological features. Although we do not engage in such calculations consciously or explicitly, we engage in them unconsciously and implicitly, and very, very quickly. After all, computers perform derivations in a flash, and, surely, the brain is just a very complex biological computer! This manoeuvre involves a false conception of understanding as a mental act, activity or process, a conception which runs like a canker through all such generative theories. We shall return to this. It also commits the homunculus fallacy47

in a flagrant manner. We shall pass over this ab su rdity with the simple remarks that what the brain does and what a person does are two different things, that brains do not understand anything, and that if brains per­ formed 'derivations', they would have to communicate the results of the derivation to the persons whose brains they are. This would require that both brain and person speak a common language. But the brain neither speaks nor understands any language. It would also require that the person understand the brain's communication, for which, ex hypothesi, he would need a further brain to calculate its meaning from the meanings of its constituent symbols and their mode of combination. Could our qualms about tacit knowledge not be put to rest by a terminological adjustment? Chomsky, confronted by objections to his misuse of the term 'know', introduces as a 'technical term' the expression 'cognize':

342

47 A. J. P. Kenny, 'The Homunculus Fallacy', in Interpretations of Life and Mind, ed. M. Grene (Routledge & Kegan Paul, London, 1 971).

343

Let us say that if a speaker knows the language L, then he cognizes L. Furthermore he cognizes the linguistic facts that he knows (in any uncontroversial sense of 'know') and he cognizes the principles and rules of his internalized grammar, both those that might be brought to awareness and those that are forever hidden from consciousness.48

Indeed, it turns out that one cognizes principles of universal grammar, e.g. that transformations apply in cyclic ordering and obey SSC, that initial phrase markers and surface structures contribute to semantic interpreta­ tion, and that transformations are structure-dependent. In short, there is more cognized in Chomsky'S philosophy than is dreamt of in heaven and earth. Is the matter a mere terminological disagreement? Chomsky suggests as much: It is unclear that more than terminology is at stake here. [One] might choose to abandon the terms 'knowledge' and even 'knowledge oflanguage' (if some find that offensive), while noting that there is little warrant in ordinary usage for these decisions. If so, he will speak of acquiring, cognizing and competence, instead of learning, knowing and knowledge. As long as we are clear about what we are doing, either approach seems to me quite all right. 'Provided we agree about the thing, it is needless to dispute about the terms' (Hume).49 48 Chomsky, Reflections on Language, pp. 164f., d. Rules and Representations (Blackwell, London, 1 980), pp. 69f. 49 Ibid., p. 166.

344

The Generative Theory of Understanding

Indeed, no one need suggest that we abandon the term 'knowing English'. It is unproblematic, and we all know how to use it. One may indeed know a language more or less well. The Bard's mastery of English was unique, and in addition he knew a little Latin, and less Greek. But did he know language ? This phrase is not one we are all familiar with, and it requires explanation. When a person knows something he can typically tell one what he knows or display what he knows (and not merely that he knows) when asked, or at least recognize what he knows when it is specified (if he found it difficult to articulate). But, of course, none of this applies to knowing principles of Chomsky's theoretical linguistics or of truth-conditional theories of meaning. These objects of putative knowledge an ordinary speaker cannot rehearse on demand; he cannot immediately recognize a theorist's rehearsal as a correct articulation of what he was attempting unsuccessfully to express. So what is meant by saying that he 'tacitly knows' these theoretical mysteries? Invoking Hume's confusions in support of modern muddles hardly shifts the onus of proof. Unless the disputed terms are clarified, we do not even know what 'thing' it is about which we are supposed to agree. At least some cases of using the phrase 'tacit knowledge' can be clarified by elaborating the criteria for tacitly knowing (say, Chopin trills) as opposed to explicitly knowing or being ignorant. But what exactly dif­ ferentiates cognizing from knowing? Chomsky insists that: I don't think that 'cognize' is very far from 'know' where the latter term is moderately clear, but this seems to me a relatively minor issue, similar to the question whether the terms ' force' and 'mass' in physics depart from their con­ ventional sense (as they obviously do).50

But physicists have the decency to give very precise definitions of the terms ' force' and 'mass', and their divergence from the non-technical use of these words is by no means a relatively minor, but an absolutely crucial (if perspicuous) issue. The use of 'cognize' (or 'tacitly know') is only 'explained' to the extent that it is said to be just like 'know', except that one who only cognizes cannot tell one what he cognizes, cannot display the object of his cognizing, does not recognize what he cognizes when told, never (apparently) forgets what he cognizes (but never remembers it either), has never learnt it and could not teach it, and so on. In short, 50 Chomsky, Rules and Representations, p. 70.

The Generative Theory of Understanding

345

cognizing is just like knowing, except that it is totally differe nt in all respects. This is a travesty of the term 'know', of the introdu ction of technical terms in science, and of respectable reasoning. If there is any point in using language at all it is that a word is taken to stand for a particular fact or idea and not for other facts or ideas. I might claim to be able to fly . . . Lo, I say, I am flying. But you are not propelling yourself about while suspended in the air, someone may point out. Ah no, I reply, that is no longer considered the proper concern of people who can fly. In fact, it is frowned upon. Nowadays, a flyer never leaves the ground and wouldn't know how. I see, says my baffled inter­ locutor, so when you say you can fly you are using the word in a purely private sense. I see I have made myself clear, I say. Then, says this chap in some relief, you cannot actually fly after all? On the contrary, I say, I have just told you I can. 51

This redefinition of the word 'fly' is no more intelligible than Chomsky's explanation of 'cognize'. Until we are told what it is to cognize but not know something, until we are given criteria for saying of someone that he cognizes something, and until it is explained to us how one can 'utilize' what one cognizes (but does not know) in understanding and speaking a language, we need pay no more attention to claims about cognizing theories of meaning or grammars than to the anatomy of borogoves.

6 Understanding new sentences It has seemed to many theorists that the most powerful argument in favour of representing 'the workings of language' in the form of a calculus is precisely that it apparently can explain how it is possible for us to under­ stand sentences we have never heard before. This 'creative' capacity of human beings is commonly thought to be at the root of much that is essentially and uniquely characteristic of our species. Does not the explanatory power of the very idea of a systematic theory of meaning of the kind in question justify it in the face of any criticism, including the arguments thus far rehearsed? Chomsky evidently believed that repudia­ tion of his theories would require one to embrace behaviourism. One might doubt whether Chomsky's ghostly machine in the mind is the only alternative to Skinner's clockwork. But even if one did reject this crude dichotomy, one might think that any theory, or at least the programme for such a theory, which held out the hope of explaining our ability to 5 1 T. Stoppard, Travesties (Faber, London, 1 975), pp. 38£.

346

The Generative Theory of Understanding

The Generative Theory of Understanding

understand new sentences, was to be preferred to any account which left so fundamental a feature of human nature shrouded in mystery. Accordingly one will be prone to view objections as provisional difficulties, the cost of desirable, bold hypothesizing, to be paid for later when the theory is further developed. This response is quite mistaken. There is, we suggest, no mystery here, a fortiori no need for theories to render something comprehensible. Rather, 'we interpret the enigma created by our mis­ understanding as the enigma of an incomprehensible process'.52 What is needed is more careful scrutiny of the concept of understanding (not theories about phenomena or processes which, we think, must occur since otherwise we would not be able to understand). Then the question 'How is it possible to understand new sentences ?' will lose its aura, and its muddled presuppositions will be brought to light. What is the apparent problem? Let us rehearse it yet again, in the favoured manner. We are, as language learners, exposed to a limited range of sentences. Somehow, on the basis of this exposure, we acquire the ability to understand infinitely many sentences. We can, in this way, do things we have never learnt, employ sentences we have never heard before and respond coherently to novel sentences. As Chomsky remarks, 'It is immediately obvious that the data available to the child is quite limited the number of seconds in his lifetime is trivially small as compared with the range of sentences that he can immediately understand . . .' 53 The infinite range of our understanding must be explained. The only way to do so appears to be by reference to a theoretical grammar ( 'a function in intension' ) or theory of meaning for a language which is 'tacitly known' or 'cognized' by a speaker. This knowledge or cognition can explain the processes involved in understanding. Before jumping to the customary conclusion, we should re-examine the seemingly profound question 'How is it possible for us to understand new sentences ?'. It is by no means perspicuous. Is it a philosophical question? With its Kantian echo it seems to be one. Or is it an empirical one, to be answered by linguists, who themselves insist that they are empirical scientists ? Does it differ from the question 'How do we understand new sentences?', and does that question make sense? We shall examine three central issues: ( 1 ) the concept of understanding; (2) the kind of possibility

in question; and (3) the relevance of the novelty of the object of under­ standing. The question 'How is it possible to understand new sentences?' appears to be akin to any question that asks for the methods for the performance of an act or activity. It looks like the perfectly clear question 'How is it possible to start a motor car without an ignition key?'. It is clear that theorists of language typically conceive of it thus. Hence they answer it on the pattern of answers to questions of how one does something. In particular, they explain that understanding is possible in virtue of tacit knowledge of a grammatical theory or theory of meaning. For, as we have seen, it is held that 'a process of derivation of some kind is involved in the understanding of a sentence'. 54 The speaker's understanding of a sentence is held to involve an instantaneous determination of its generation by the rules of the theory ('an adult can instantaneously determine whether (and if so, how) a particular item is generated by this mechanism' ).55 Or, if understanding a sentence is conceived not as an act or activity but as a state, then the question about its possibility is conceived as a question about the mental processes or activities that bring about this state: 'What a speaker does when he understands or produces an utterance must include at least the implicit analysing of its syntactic structure . . . [and the assigning of] interpretations to sentences on the basis of his knowledge of the meanings of their parts.'56 But understanding is neither an experience one undergoes nor an act one performs, although there are commonly many experiences which accom­ pany understanding an utterance. No inner act or activity is either neces­ sary or sufficient for understanding, for performance of any such mental acts or activities is compatible with gross misunderstanding or failure of understanding. Indeed, we would not deny a person's understanding a given utterance, where this is manifest in his behaviour and speech, on the grounds of the absence of some correlative mental act, activity, or process. A fortiori unconscious inner processes, even if there were any evidence for such implicit analysings, derivings or assignings, neither constitute nor are either necessary or sufficient for understanding. Not only is understanding not an act, activity, or process one engages in, and of which one might ask 'How does one do it?', it is not a state of mind brought about by an antecedent act, activity or process of the mind. The

52 L. Wittgenstein, Philosophical Grammar, ed. Rush Rhees (Blackwell, Oxford, 1974), p. 1 5 5 .

5 3 N. Chomsky, Language and Mind, enlarged edition (Harcourt Brace Jovanovich, Inc., USA, 1 972), p. 1 13.

3 47

54 M. A. E. Dummett, 'What is a Theory of Meaning?', in Mind and Language, ed. S. Guttenplan (Clarendon Press, Oxford, 1975), p. 1 12. 55 N. Chomsky, 'Review of Skinner's Verbal Behaviour', Language, 35 ( 1 959), pp. 57f. 5. Fodor and Katz, 'What's wrong with the Philosophy of Language?', p. 282.

348

The Generative Theory of Understanding

The Generative Theory of Understanding

reason for this is not that mental acts or activities are never engaged in prior to, and as a route to understanding something (as in recapitulating the steps in a proof in the endeavour to understand a theorem), it is rather that understanding is not a state of mind at all. 57 There is no such thing as being in a state of understanding a sentence, an utterance or a language. One may understand something from a certain time (the time at which one learnt it or read it with understanding) for a certain time (as long as one passes 'tests' of understanding, i.e. satisfies criteria of understanding). But though I may have understood, been able to speak, French from the age of fourteen to my mid-thirties (by which time it had become so rusty as to be useless), it would be absurd to say that I understood it continuously for twenty years, and no less absurd to say that I understood it intermittently. Understanding is subject to degrees in many cases, since I may understand a lecture, an hypothesis, a statement, more or less well, but not more or less intensely. If my concentration is interrupted I may not understand what I am concentrating on, but given that I understand it, my understanding cannot be interrupted. When I fall asleep my excitement at having under­ stood a difficult problem abates, but my understanding does not cease. One may indeed suddenly cease to understand a problem, and one may equally suddenly regain one's understanding of it (the solution may have a Gestalt-like quality). But this is not like the abating of a pain under the impact of an analgesic, and its subsequent resumption when the drug wears off. Rather is it like suddenly forgetting and subsequently recollect­ ing. I may indeed forget what something (e.g. a statement of a theorem) means, and then I no longer understand it. But forgetting something previously understood is not a fall from a state of intellectual grace. (Yet note that although I may have understood what someone said when he said it, it would be wrong to say that when, half an hour later, I have forgotten what he said, I no longer understand what he said.) 58 This negative analysis is by no means fruitless. By reference to it the intellectual landscape may take on new aspects. Note, for example, the striking fact that theorists who present a generative account of under­ standing sentences typically focus upon understanding sentences uttered by someone else. This tendency is not accidental. For understanding what one hears or reads lends itself to the favoured picture of understanding as an act, activity, or process, or as a state produced by such antecedent acts,

actlvltleS, or processes. One hears a sequence of noises, it seems, and recognizes them as words whose meanings one knows. 'As quick as a flash' one derives the meaning of the sentence from the known meanings of the words and their mode of combination by using the principles of the tacitly known grammar or theory of meaning of the language. Then, having carried out this instantaneous calculation, further aided perhaps by a theory of force or principles of pragmatics, one understands what one heard. Just how misguided this picture is becomes evident if, bearing in mind that understanding is not an act, activity or state, we view the matter from the perspective of a normal speaker who understands the sentences he himself utters and knows what he is saying by using them. The hearer understands a sentence, the theorist is inclined to say, when he has heard it and carried out the appropriate derivation. But when does a speaker understand a sentence he utters, according to this account? Before he utters it? But how is it possible for him to understand what he says before he says it? Indeed, what is there to understand before the sentence is spoken? Is it that he speaks it to himself quickly before he speaks it aloud? But that idea merely generates an infinite regress, for the question of when he understands the sentence he silently says to himself now arises. If only after he says it to himself, does he then say it to himself without under­ standing it, as it were, to see what sense it will make? This is absurd. But if before he says it to himself, then how is that possible? Antecedent silent soliloquy is an empty move. Does it then follow that a speaker does not understand what he says aloud until after he has said it, that he must wait to hear what he says before he can know what he means? This too is absurd. Here the favoured picture of understanding clashes violently with its application. Of course a speaker typically understands what he says. But his understanding is not something that occurs before, simultaneously with, or after his utterance, any more than being able to play chess occurs before or after moving a chess-piece in a game. Understanding is akin to an ability,s9 rather than to a state, act, or activity. It is an aspect of our concept of understanding that a person who understands an utterance can typically do certain things, e.g. explain what it means, respond to it appropriately, paraphrase it. His understanding manifests itself in the exercises of such abilities. And it is the overt exercises of these abilities that constitute the grounds upon which we legitimately ascribe understanding to him. Such behavioural manifestations constitute

57 For a brief characterization of states, see above, p. 279f. 58 For a more detailed account of understanding, see G. P. Baker and P. M. S. Hacker, Wittgenstein: Understanding and Meaning, (Blackwell, Oxford, 1 980), pp. 595ff.

349

5. Although it would be misleading to say that it is an ability, d. Baker and Hacker, Wittgenstein: Understanding and Meaning, pp. 6 17ff.

The Generative Theory of Understanding

The Generative Theory of Understanding

the criteria justifying the ascription of understanding to a person. They do not entail understanding, since a criterion may obtain, yet be defeated in the circumstances. Yet the behavioural manifestations of understanding are not inductive evidence of understanding either, since that would suppose that we had a concept of understanding enabling us to identify cases of understanding independently of behaviour, which we could then inductively correlate with the behaviour.60 Understanding is not identical with the behaviour manifesting it, nor is it an inner state or process from which that behaviour flows. Its kinship with an ability is here evident. The power or ability of a car to go at 70 m.p.h. is not identical with its going at 70 m.p.h., but nor is it an inner state or activity of the car. An ability is no more identical with what might be called61 its vehicle or the structure of its vehicle than it is with its manifestations or exercise. One cannot find the car's horsepower under its bonnet, or a person's understanding in his brain. Consequently, the question of how it is possible for a person to under­ stand new sentences boils down to the question of how it is possible for a person to be able to do those things which manifest understanding, namely react to, use and explain the meanings of new sentences. This question is still far from clear, but it is clear that it cannot be correctly understood as a question about the means or methods of performance of a special mental act or activity, viz. understanding. Of an ability one cannot ask 'How does one do it?'. A fortiori it makes no sense to ask how it is possible to do it. There are no methods of understanding, any more than there are methods, ways or means of being able to do something. There are methods of learning how to do something, ways of bringing it about that someone is able to do something or other, i.e. methods for the acquisition of an ability, but no methods of 'doing' the ability. Since abilities are not 'done', and since understanding a sentence or utterance bears a conceptual kinship to an ability, it is not surprising that the question 'How is it possible to understand a new sentence?' should occasion a sense of mystery. For interpreted in one way, it gravitates towards categorial confusion. One does not understand a sentence any how.62 Of course, there are circumstances in which the question 'How can

you understand ?' is in order. If a speech is given amidst hubbub, one might ask someone how he can understand what is being said in the uproar. The answer might be that he lipreads; but lipreading is not a method of understanding the meanings of sentences but one way (like hearing) of discerning what a speaker is saying. The question here concerns circum­ stantial possibility, namely given that most people cannot understand utterances in such a noise (since they cannot hear the words spoken), how it is that this person is not prevented by these circumstances from under­ standing, how is it that he can tell us what is said, explain it, paraphrase it, etc. This kind of question is not at issue in the case of a speaker's under­ standing sentences or utterances in his native tongue in optimal circum­ stances. A different kind of case arises when someone is seen to exercise a complex ability which has not apparently been learnt and yet is obviously not innate. Thus a hearer might explain that he can understand a lecture in Spanish, even though he has never learnt Spanish, because he knows Italian and Portuguese, and can guess most of what is being said on that basis. This explains the possession of an unlearnt ability as a consequence of possession of other learnt abilities (as one might explain one's ability to play squash (after a fashion) by reference to one's having already mastered tennis and badminton). But neither learning Italian and Portuguese nor knowing Italian and Portuguese are means or methods of understanding a Spanish lecture, even though this person's understanding it can be ex­ plained by reference to his mastery of Italian and Portuguese. Again, this kind of question is not at issue when one asks how it is possible for one to understand sentences of one's mother tongue. Given that understanding is akin to an ability, the question as to its possibility seems to amount to the question of how it is possible for a person to do those things that are manifestations of understanding. How should this be interpreted? We can rule out any question of methods of doing those things (how to shut the door in compliance with an order to shut it). Is it then a question about empirical causes and psychological or neurological underpinnings of our linguistic ability? We can construe it in various ways. It may be taken as a question about the structure of the vehicle of our linguistic ability; or about the mode of acquisition of the ability; or about the prerequisites for the possession of the ability. Explanations of powers and abilities are characteristically given, in the advanced sciences, by specification of the physical structure of the bearer of the ability (e.g. the magnetic attraction of a magnet or corrosive power of an acid are explained by reference to, although not reduced to, mole­ cular structures). If one wishes to speak of the vehicle of an ability, one

350

60 Bear in mind that the first-person utterance 'I understand' is not typically a judgment resting on evidence nor is it an identification of a mental act, activity or state. 61 Cf. A. J. P. Kenny, Will, Freedom and Power (Blackwell, Oxford, 1 975), p. 10. 62 Although there is, of course, a perfectly straightforward use for 'How do you under­ stand this sentence?', namely 'How do you take it, what do you understand by it?'

351

The Generative Theory of Understanding

The Generative Theory of Understanding

might be inclined, with some qualification and risk of confusion, 63 to claim that the vehicle of a person's understanding is his brain. Neurophysiology may, one day, give a description of the neural structure of the cerebral cortex which will contribute to our understanding of the physiological foundations of our linguistic and cognitive skills. Of such possible explanations three points should be noted. First, such explanations are empirical and contingent. They will, therefore, not reveal any internal relations between understanding and language which it is the business of philosophy to investigate. Nor will they reveal any structural properties of languages which it is the business of grammarians to study. Secondly, there is no a priori guarantee that significant or extensive explanations will be forthcoming. There may simply not be any sharp and determinate correla­ tions between cerebral organization and the highly diffuse and complex abilities and dispositions and their indefinite and multiple forms of exercise which manifest understanding sentences and utterances. Thirdly, such an explanation would not be normative, i.e. it would not display linguistic behaviour as a case of rule-guided conduct. One could not see from the cerebral structure that is (according to the supposed theory) essential for linguistic mastery in all its diversity which rules are being followed, or discern what counts as correct compliance with them. Normative phenomena are public and social, not private and neural. If the question of how it is possible to understand new sentences is concerned with cognitive prerequisites of understanding, we might construe it as a psychological question. For it might be a question about the mode of acquisition of linguistic skills, or some facet of mastery of a language. To that extent it would aim to elicit the psychological or pedagogical prerequisites for their possession. Taken thus the question is far too crude for any answer. Duly refined, and broken down into numerous manageable questions, it would perhaps be interesting and fruitful. But be that as it may, it is a matter for empirical investigations in experimental psychology. These issues, however, are wholly undeveloped and bedevilled by misguided theories concocted by linguists and philo­ sophers which stand between the psychologist (learning theorist) and careful observation of the phenomena. The question so construed is also philosophically irrelevant. Finally, we may take the question analytically. For it can be construed as a question about what a person must know in order to understand new sentences. The general form of the answer to this is neither mysterious nor

startling: one must know whatever is requisite to satisfy the criteria for understanding new sentences, i.e. know whatever is necessary for the performance of those acts or for those reactions and responses the ability to perform which is constitutive of understanding. For how else can it be shown that one must know something in order to understand a given new sentence ? One can only show this by demonstrating that the manifestation of understanding is internally related to possession of such knowledge. Hence, altogether unsurprisingly, one normally has to know what the constituent words of the given sentence mean, since in explaining what the sentence means one typically also explains what its constituents mean in that context. But, equally unsurprisingly, one does not have to know any depth-grammar or 'logical form', since manifestation of understanding of a new sentence is wholly independent of manifestation of knowledge of theoretical linguistics or philosophy of language. Three things now become evident. First, the novelty of the sentences one understands is a red herring. It seemed relevant insofar as it was an objection to crude behaviourist theories of understanding that they could not make sense of the fact that we can understand new sentences. But stimulus-response theories cannot 'explain' understanding previously heard sentences any better. The novelty of a sentence is immaterial. For the grounds for attributing to a person an understanding of some sentence are not different when it is a new sentence from what they are when it is familiar. Indeed, we commonly do not know, when we judge that someone understands it, whether it is novel or familiar to him. Secondly, the question of the number of sentences we can understand is profoundly misleading. Contrasting the 'trivially small' number of seconds of the child's life with the vast number of sentences he can understand encourages confusion. How very clever of the child to learn so much in so little time ! We think that in some sense the child 'already' understands an infinity of sentences now, even though he has only heard a very small range of sentences.64 If he can understand, we are inclined to argue, then in some sense he does understand already. But if he already understands, we confusedly think, then the infinity of sentences must be, as it were, stored up in the child's mind, if not in the form of encoded sentences, then in the form of a grammar, a lexicon, and a decoding device. These myths stem from treating the illegitimate question 'How many sentences can the child understand?' as if it were parallel to 'How many

352

63 It is unclear, for example, why the person himself is notthe vehicle of his understanding.

353

64 As Zeno might have thought that the toddler has traversed an infinite space, on the ground that his few steps contain an infinity of spaces!

354

The Generative Theory of Understanding

sentences has the child uttered?'.65 But it is not. Understanding shares the open-endedness characteristic of abilities. It is of the nature of many abilities that in acquiring them one becomes able to do many things not hitherto done. Many human skills are plastic. Learning to draw or to paint, to throw pots on a wheel, to act or mime, and so ort, all involve an open-endedness which is misrepresented if it is held to be especially mysterious or baffling. It would be absurd to be amazed at the vast number of possible pictures the artist can paint, the huge number of possible pots the potter can throw, the endless number of potential roles the actor can act. For not only is the plasticity of the skill not mysterious, the idea of counting the number of possible pictures that one could paint, and com­ paring it with the 'trivially small' number of seconds of one's life is silly. What would really be mysterious is a person's only being able to under­ stand (and speak with understanding) 7,568 sentences, and not a single one more. That would be baffling! Thirdly, there is no unique thing that a person must know in order to satisfy the criteria for understanding some arbitrary sentence. He must, indeed, use it correctly, react to it appropriately and explain it cogently. If he does these, then other things being equal, we attribute understanding to him. But there is no general mechanism for 'derivation' of the meaning of the sentence from its constituents which is pertinent to the criteria of understanding. There is not, in general, a privileged form of explanation of what a sentence means which requires the use of any special knowledge in order that a person should justifiably be said to understand a sentence. In particular, knowledge of a theory of meaning for a language or of a transformational-generative grammar does not have to be displayed in explaining what a sentence means in order to satisfy criteria of under­ standing. And conversely, Chomsky'S knowledge of his grammar and his use of it in explaining a line in Shakespeare, say, is perfectly compatible with his grossly misunderstanding this line. Our ability to understand sentences is not an ability to make lightning­ quick calculations in which we derive the meaning of a sentence from the meanings of its constituents and their mode of combination. We typically grasp or 'take in' the meaning of a sentence or utterance as a whole, and not by means of any computational techniques. It is quite false to suggest that we hear nothing but noises or read nothing but marks on paper, which 65 Cf. Wittgenstein's Lecturers on the Foundations of Mathematics, Cambridge, 1 939, ed. Cora Diamond (Harvester Press, Sussex, 1 976), pp. 3 1 ff., for a parallel discussion of the question of how many numerals one has learnt to write down.

,

The Generative Theory of Understanding

355

dead signs we have to recognize and process, endowing them with an appropriate meaning, and then computing the meaning of the whole from the meanings of the parts. This is of a piece with the suggestion that what we see are just colours and shapes (not chairs and tables, trees and buildings) which we then have to process and 'interpret' to yield our 'picture' of the objective world. This ancient myth of the 'given' and of what the mind, contributing structure from its own resources, makes of it is not rendered respectable by being dressed up in late twentieth century garb. It makes sense to ask me how I know that when you said 'Give me a drink' you wanted a gin and tonic, but not to ask how I knew that you wanted a drink. This is parallel to the fact that it makes sense to ask how I know that a seventeenth century Dutch still-life of flowers in a vase represents the fleetingness and vanity of life, but not to ask how I know that it is a still-life of flowers in a vase.66 In this respect understanding sentences is not unlike understanding genre paintings. Indeed, a language theorist may be equally impressed by our ability to understand an 'infinite' number of paintings (putting historical paintings aside). He may even insist that the possibility of this understanding must be explained in generative terms. The analogy is perhaps worth pursuing, for it may help relieve the manifold pressures that push one into embracing complex grammatical theories or philosophical theories of meaning as essential underpinnings for an explanation of understanding. In the case of sentence and painting alike the whole is not the mere sum of its parts, but of its parts and their mode of combination. Similarly, in both cases the meaning of a part depends on its juxtaposition with other parts, and with features of the whole which are neither parts nor relationships of parts (e.g. chiaroscuro) . In both cases there are rules of representation which are presupposed, and also more complex theories of representation which may be known (e.g. rules of fixed point perspective) but which need not be known for the understanding of the painting. Strikingly, in both cases the identification of the unit of significance in the whole is problematic. For although one may say that the painting is composed of parts, it is impossible to say, in advance of specific purposes, what is to count as a part - a line, a patch of colour, a shape, a determinate representation (and each of these is in turn problematic). In the case of 66 Of course, in both cases a psychologist may investigate what aural or visual elements are indispensable 'cues', in the sense that if they are removed or interrupted in some way, I will not be able to say what the painting is of, or what the utterance meant. It does not, however, follow that such 'cues' are my grounds for 'recognition'. There is here typically no question of recognition, no inference, a fortiori no ground of judgment.

356

The Generative Theory of Understanding

, "

paintings, as with language, the conventions of representation must be learnt. The most important analogy lies at the level of understanding. We take in a painting as a whole. We do not understand it as, e.g., a painting of a mountain landscape in a rainstorm with cattle drinking at a brook, by calculating or constructing this correct 'interpretation' from our grasp of (the meaning of) the parts and their mode of combination. Of course, we can anatomize the painting, identify the cattle, brook, falling rain, etc., even identify the various brush strokes and impasto that create the image of a lowing cow. But it does not follow that our understanding of the whole rests on, has as its grounds the antecedent anatomization of the picture and the derivation of the meaning of the whole from the meanings of its parts. Nor, indeed, does it mean that we can anatomize a painting into elementary constituents identifiable as such-and-such representations independently of the picture as a whole. There is no specially problematic question or deep mystery about our capacity to recognize a painting of something that we have never seen (e.g. a knight killing a dragon) as opposed to a painting of something familIar (e.g. a vase of flowers). In neither case do we need to know any complex theory of pictorial representation. (Of course, there is ample room for psychological and physiological investigation into perception and recognition.) There is no deep question of how it is possible for us to understand sentences we have never heard before, only deep confusions about the concept of understanding, which we then mistake for mysteries about the phenomena of understanding. Consequently, the alleged capacity to answer this question is no vindication of the idea of a theory of meaning for a language, or of a theory of grammar tacitly known by every speaker. The seeds sown by the Tractatus and Frege have indeed germinated. The plants that have grown from them infest the intellectual landscape. But despite their size, they are barren.

7 Residual business: the minimalist enterprise We turn finally to the minimalist enterprise pursued by some philosophers. They construct 'theories of meaning for a natural language', but hasten to add that these theories are not, or not directly, of any psychological moment. They do not explain (although they might be invoked in explain­ ing) how language is learnt or how linguistic skills evolve. Their analytic role has been variously characterized. On one view a theory of meaning is a conceptual clarification of what a language is. This being achieved, the

The Generative Theory of Understanding

357

results can be handed over to a psychologist who can investigate, if he wishes, the 'mechanisms' whereby 'linguistic input is processed' to yield understanding.67 According to a different conception,68 a theory of mean­ ing specifies an array of axioms and theorems together with principles such that, if someone knew them, he would be able to speak the language. This is alleged to provide a clarification of 'what is involved' in speaking and understanding a language, even though it is not contended that actual speakers really know any such theory.69 A third variation on this apparently modest theme is the suggestion that a theory of meaning is 'a theoretical representation of a practical ability'. It represents or charac­ terizes, in a systematic manner, what it is that a person who has mastered a language, acquired the practical ability of which understanding a language consists, can do. H our criticisms of the very project of constructing a theory of meaning are correct, if the necessary distinction between sense and force is in­ coherent, if the various ideas of the truth-conditions of a sentence are all muddled, if the representation of a language as a calculus of meaning-rules is a misrepresentation, then this way of explaining what the mastery of a language consists in is foreclosed. We shall, however, suspend our disbelief and examine the feasibility of one of these projects afresh. We shall not dwell on the first proposal. It is no less absurd to suppose that an elaborate axiomatized theory provides a conceptual clarification of what a language is than to maintain that the logical calculus of Principia Mathematica first revealed to us the true meanings of our familiar number­ words. Even if the theorist could deliver a complete T-theory for English, that would not show that English is a calculus of T-sentences, any more than a Mercator projection shows that the Earth is flat. Proficient speakers of English use names of languages such as 'English', 'French', 'Chinese' perfectly correctly, but they do not mean, by these names, to signify axiomatized theories (still awaiting construction!) for the derivation of se�tence-meanings from a set of axioms (or specifications of word67 This in effect hands the philosopher's dirty linen over to the cognitive psychologist. When the latter engages in pseudo-science of the kind previously discussed, and is criticized for so doing, the theorist of meaning will typically tum, with deceptively innocent tones, and insist that the psychologist is an empirical scientist whose theory-building endeavours lie beyond the purview of conceptual analysis and criticism. 68 D. Davidson, 'Reply to Foster', in Truth and Meaning, ed. G. Evans and J. McDowell, ( Clarendon Press, Oxford, 1 976), pp. 33f. 69 Cf. M. A. E. Dummett, 'Objections to Chomsky', London Review of Books, 1-14 October, 1 9 8 1 .

358

The Generative Theory of Understanding

meanings). Rather, we use 'a language' to designate a complex normative practice governed by motleys of rules familiar to and acknowledged by competent speakers. There is, of course, more to be said about this first proposal. This will be done when the third variant is subjected to critical . scrutiny. We shall pass over the second suggestion almost equally quickly. It is typically propounded as an empirical thesis about languages. Thus con­ strued, it is objectionable. In the first place, knowing the axioms and being able to derive the theorems of a theory of meaning for a natural language would not, in any sense, qualify one as a speaker, any more than knowing the rules of cricket would render one a qualified batsman or bowler. Mastery of a language is a skill, and it is not a skill which manifests itself in performing derivations of T-sentences. Mere knowledge of such a theory would leave one, in Harris's phrase, a communicational cripple (and this would not be because one had mastered the semantics of a language, but was shaky on the pragmatics). On the contrary, if we were concerned to clarify what is 'involved' in speaking English, we might more plausibly mention dictionaries and school grammars. After all, it is by using these that foreigners come to be able to speak English. If it is claimed that this enabling 'knowledge' is insufficiently systematic, inasmuch as it does not display the derivability of sentence-meanings from axioms and formation­ rules, the reply is surely that the knowledge that suffices for speaking English does not have to incorporate anything 'systematic' in this sense as a necessary component. In order to know what utterances mean, one does not have to know how to derive their meanings from anything at all. The requirement that the knowledge which would suffice to enable its possessor to speak a language take the form of a systematic calculus stems from the fallacy that only thus can one 'explain' the possibility of the 'infinity' of sentences of a language. Finally, it is altogether opaque why specification of a body of knowledge allegedly sufficient to enable one to speak a language (though evidently not necessary, since it is not contended that anyone does know the theory) should be held to provide an illuminating account of 'what is involved' in understanding a language. The thesis can, however, be taken not as an empirical claim, but as an analysis of the concept of understanding a language. If understanding a language is conceived as being able to assign to any arbitrary sentence of the language its truth-conditions, and if a theory of meaning is an axio­ matic theory which assigns to every well-formed sentence its truth­ conditions, then trivially knowledge of the theory of meaning suffices for understanding the language. On this construal, the second suggestion

, 1·,'. �:�

The Generative Theory of Understanding

359

collapses into the third. It must, however, be stressed that the theory cannot both be an empirical claim as to what knowledge suffices for understanding a given language and also an a priori claim about what understanding a language consists in. We shall focus on the third view, namely that a theory of meaning for a natural language is a 'theoretical representation of a practical ability'. According to this conception a theory of meaning 'present[s] an analysis of the complex skill which constitutes mastery of a language . . . what it is that someone who possesses that mastery is able to do'.70 It is argued that understanding is a 'practical ability' and that abilities are characterized by specification of what it is that they are abilities to do. It is, of course, recognized that among the things a speaker of a language can do is to perform a myriad of speech-acts. These are given a 'theoretical representa­ tion' by means of the theory of force. ' [A]gainst such a background . . . it makes sense to say that to know the meaning of a sentence is to know the condition for its truth'.71 Waiving objections to the intelligibility of such a theory of force, we focus here on whether understanding a language is a 'practical ability' intelligibly 'represented' by means of a theory of mean­ ing, in particular whether this ability is correctly characterized in terms of an axiomatic theory for derivation of truth-conditions of sentences from stipulated axioms. The philosopher's primary move is to contend that to understand a sentence is to assign (or to be able to assign (?) ) to the sentence a set of truth-conditions. But there is no such thing as understanding a single sentence of a language and no other (although one might know the meaning of a single Chinese sentence). To understand a sentence is to understand a language. A language is, it is held, an infinite set of sentences. Understanding a language is a remarkable capacity precisely because it involves the ability to comprehend an infinity of sentences. But now, it is argued, we cannot characterize the mastery of a (given) language by a complete enumeration correlating each of its 'infinity' of sentences with its truth-conditions, even though so pairing sentences with truth-conditions 70 Dummett, 'What is a Theory of Meaning? (II)" p. 70; as noted above (p. 325) Dummett's account is marred by constant intrusion of remarks about implicit knowledge, implicit derivations of theorems from axioms, coupled with insistence that the account involves no psychological hypotheses, even though this knowledge is said to issue in a general ability to speak (Ibid., p. 71), that implicit grasp of general principles issues in recognitional capacities to determine syntactic well-formedness, that the speaker derives the meanings of sentences from implicit knowledge of axioms. We shall disregard this muddle. 7 1 Ibid., p. 73.

360

The Generative Theory of Understanding

would be, sub specie aeternitatis, a correct characterization of what under­ standing this language is. Rather, as mere finite mortals, we must formu­ late general laws from which, for an arbitrary well-formed sentence, we can systematically calculate its truth-conditions. Mastery of a language will be characterized correctly as an ability to derive the logical con­ sequences of such laws, viz. the meanings of sentences of the language. Or, on the assumption that derivations take care of themselves, it will be characterized as 'knowledge' of these general laws, or of the ' function in intension' which generates an infinite set of correlations of sentences with their truth-conditions. Knowing a language (mastery of a language) is identified not merely with knowledge of the axioms of an axiomatized theory, but rather with this together with knowledge of general rules capable of iterated applica­ tion. Assigning truth-conditions to an arbitrary sentence is thought to parallel stating the answer to an arbitrary multiplication problem. In the latter case, a recursion formula such as (a + 1 ) x b a x b + b, together with the axiom 0 x 1 = 0, determines what integer is paired with any pair of integers as their product. Hence mastery of multiplication is identified with knowledge of this recursion formula (and the axiom) . Similarly, knowledge of a language is conceived as knowledge of axioms and recur­ sion formulae which correlate every well-formed sentence with its truth­ conditions. Accordingly, a theory of meaning for a language will provide a theoretical representation of a practical ability (a characterization of what understanding a language consists in), just as recursion formulae in arith­ metic constitute a theoretical representation of a practical ability (know­ ing arithmetic). The number of prestidigitations in this brief account is very great. To see how the conjuring trick is effected, we shall have to examine it closely, as it were in slow motion. Prior to detailed scrutiny, however, two crucial points about its general contours must be stressed. First, this reasoning provides no support whatever for a generative theory of understanding or a compositional theory of meaning. Rather, it presupposes some such theory. Unless the meaning of an arbitrary sentence could in principle be calculated systematically from its con­ stituents and its structure, there would be no such thing as a recursive specification of the truth-conditions of the infinity of sentences in a language, and hence the analogy between understanding a language and the ability to multiply would collapse at a vital point. Without some independent argument, we have no assurance that it is even intelligible to speak of such a theoretical representation of the ability to speak English, =

,

The Generative Theory of Understanding

361

French or Chinese. The minimalist enterprise rests on a determination to cast natural languages into the forms of formal calculi, and hence it makes sense only if languages satisfy this Procrustean demand. But, as we have shown by a plethora of arguments, they do not. Secondly, although knowledge of axioms and recursion formulae is invoked in explaining the notion of a theoretical representation of a practical ability, it plays no essential role in the reasoning. This is important. It justifies the claim that this characterization of understanding a language is rightly classified as a minimalist account. This theoretical knowledge plays no part in any explanation of how somebody under­ stands a sentence (the subject-matter of the full-blown explanatory enter­ prise) but rather concerns only the circumscription of what he understands and a clarification of what his understanding a language 'consists in'. Furthermore, the subsidiary role of knowledge circumvents a problem. It seems that characterizing an ability in terms of knowledge of theoretical principles is altogether different from characterizing it in terms of specify­ ing what constitutes its exercise (i.e. what it is an ability to do). But this issue can be by-passed if the exercises of the ability can be characterized directly in terms of the principles themselves. This is held to be the case with the recursion formula for multiplication, and by analogy, with the principles of a theory of meaning. The formula itself is alleged to generate the correct answer to every possible multiplication problem. But it is the giving of these answers which constitutes the totality of the exercises ofthe ability to multiply. Hence the recursion formula itself seemingly circum­ scribes precisely and fully what somebody who has mastered multiplica­ tion can do, and it does so independently of any characterization of this ability in terms of theoretical knowledge (even if, under certain idealizing suppositions, that might also be illuminating). The pairing of each sentence of a language with its truth-conditions is supposed to flow from a theory of meaning in the same direct way, making no essential detour via the concept of theoretical knowledge. And it �s these pairings which are alleged to be the essential components in an analysis of the totality of the exercises of the ability to speak a language. 'Tacit knowledge', in this context, is considered to be a picturesque wrapping for such a theory of meaning. What a person who can multiply can do is just whatever some­ body who knew the recursion formula would do (assuming that he always applied it correctly and drew out its consequences, however remote). Pari passu, what a person who has mastered English can do is whatever the Ideal Speaker-Hearer who knew the entire theory of meaning for English would do.

362

The Generative Theory of Understanding

The strategy of the minimalist enterprise is clear enough. But it is unclear whether the reasoning is sound at the tactical level, and hence too whether anything intelligible is achieved. Doubts turn on three issues: the opacity of the notion of ' characterizing an ability', the interpretation of the phrase 'a full characterization of a practical ability', and the appropriateness of the analogy between the ability to speak a language and the ability to multiply integers. We shall examine each of these matters in turn. The minimalist enterprise starts from the insight that it is integral to the concept of any particular ability that it be possible to explain what counts as the manifestations of this ability. If a human agent is ascribed the ability to ride a bicycle, draw caricatures, purchase a loaf of bread, or speak English, then it must be clear what acts of his are exercises of such an ability. This can always be done by exploiting the schema: if A has the ability to cp, then A's cping is an exercise of this ability. But, of course, in many cases more illuminating accounts can be given for particular pur­ poses. We are often interested in inculcating abilities in pupils, and hence we look for ways to analyse 'complex' abilities such as playing tennis, or composing music into 'component' abilities which can be taught and practised in relative isolation from each other (e.g. playing forehands, playing backhands, volleying, serving, and scoring). There is, however, no warrant for maintaining that every human ability must admit of some such analysis into component abilities, and there is equally no presumption that any ability has a unique purpose-independent decomposition into other abilities. If either of these ideas is attached to the alleged conceptual requirement that it be possible to 'characterize' any ability in terms of what it is an ability to do, then this requirement is illegitimate (indeed, incoherent). The minimalist might abjure this requirement but none the less claim that understanding a language can be decomposed into component abili­ ties the major one of which is the ability to assign to each well-formed sentence its truth-conditions. This of course has not yet been accom­ plished; we see the 'analysis' through a glass but darkly, not face to face. The crucial question is not the grounds for this faith, but the intelligibility of what is taken on trust. As a putative analysis of a complex ability, this one makes sense only if 'assigning to a sentence its truth-conditions' is an understood expression designating a human act or activity. This is not so. The phrase is opaque, a bit of theoretical jargon standing in need of explanation. What counts as somebody's engaging in this act? Is it a private mental act known only through introspection? Or can one observe another's performing this act? And how could any act completed at a

I �

�i,

�: 1 �

t':Ii

.�

Iif ��

t

The Generative Theory of Understanding

363

particular moment be coordinated with the manifestations of someone's understanding a sentence, since these may be spread out over an indefi­ nitely long period? There is pressure here to treat assigning correct truth­ conditions to sentences as itself an ability manifested in speaking fluently, but then 'assigning to each sentence its truth-conditions' no longer charac­ terizes an ability in terms of actions which manifest this ability! The putative analysis of mastery of a language is confused and incomprehen­ sible. This negative verdict is reinforced by bringing to light muddles about the notion of completeness in characterizing practical abilities. It is obvious that an explanation of what it is to have a particular ability may be criticized because it implies an inadequate account of what constitutes the exercises of this ability. It would be a joke to say that the ability to cook well consists in the ability to make a souffle, or that the ability to play the piano is the ability to play C-major arpeggios in contrary motion. Such clarifications of abilities are clearly incomplete. This is not because they fail to meet some absolute standard of completeness. Rather, it is because they adumbrate an analysis of a complex ability, but leave out important component abilities coordinate with the ones listed. If, e.g., in clarifying what is involved in being able to play the piano, one mentions the ability to play arpeggios, it would be misleading to omit mention of the abilities to play scales and trills since these are general abilities at the same level of specificity (unless the explanation were addressed to somebody who understood that the list of component abilities was open). Whether the characterization of an ability is complete or incomplete turns on whether it leaves out component abilities which should be listed, and this depends on the purposes for which it is offered. This distinction parallels the distinc­ tion between exact and inexact descriptions, for a description is properly called inexact only relative to the purposes which it is meant to serve. 72 The concept of the completeness of the characterization of an ability is distorted by imagining that what pass for complete characterizations should be measured more rigorously against an absolute ideal of complete­ ness. As if there might be at every level of specificity a listing of acts manifesting an ability which left out no act constituting an exercise of that ability. But this ideal is chimerical (like the parallel 'ideals' of a perfectly exact description and of an explanation of meaning assigning a completely determinate sense to a word). There is typically no such thing as specifying everything that a person can do who has a particular ability, and many 72 Wittgenstein, Philosophical Investigations, § 88.

The Generative Theory of Understanding

The Generative Theory of Understanding

explanations of abilities by giving examples of component abilities (or examples of their exercises) are, and are treated as, correct and complete. One does explain what it is that persons who can cook can do without listing all the possible dishes which they can produce or all the possible activities (buttering cake tins, whipping egg-whites, etc.) which they can engage in. Moreover, if these were condemned as incomplete explana­ tions, there would be no such thing as a complete explanation of the phrase 'able to cook' because there is no possibility of listing without omission every dish that might be cooked or every action performed in cooking. But if there is no such thing as a complete explanation, then there is also no such thing as an incomplete one. If the ' full characterization' of an ability demands an explanation complete by this bogus standard, then the very idea of giving full characterization of most practical abilities is ridiculous. Mastery of a language, like many other human abilities, is both diffuse and open-ended. It is diffuse (or perhaps complex) in that no single act or activity constitutes the exercise of this ability as jumping eight metres constitutes the exercise of the ability to jump eight metres. Rather it has many facets or aspects. A competent speaker can do innumerable (but 'in­ finitely many' ?) things: make statements and assertions, give descrip­ tions or instructions, ask and answer questions, issue and obey orders, crack jokes and tell stories, wonder, beg, entreat, pray, and so on and so forth. It thus resembles the ability to play the piano. A competent pianist can play scales, trills, and arpeggios, sight-read, perform a crescendo or diminuendo, execute a phrase staccato or legato, etc. Moreover, mastery of a language (and each of its component abilities) is open-ended. There is no fixed list of compositions which a competent pianist can play (nor any closed totality of phrases) and no fixed list of tunes which a competent whistler can whistle. Similarly it would be absurd to attempt to exhaust by enumeration all of the jokes that a competent speaker could laugh at or tell, all of the instructions that he could issue, all of the questions that he could ask, etc., let alone all of the sentences that he could use or respond to. Does it follow that there must be, in the case of every open-ended ability, some formula which recursively characterizes its possible exercises? To this it might be replied that the abilities to play the piano, to whistle or compose music differ from a linguistic ability precisely because music, unlike a language, is not (in general) representational. Understanding music, it might be argued, is, pace Wittgenstein, totally unlike under­ standing English: we may, for the sake of argument, grant the objection. Nevertheless, it achieves nothing. Let us take as our examples of parallel open-ended abilities such abilities the exercise of which does result in the

production of objects that are representative, namely drawing, painting, sculpting and modelling, with the proviso that we restrict our attention to genre paintings, etc. and exclude historical representations. Let us cor­ relate with these artistic abilities the complementary abilities to under­ stand drawings, paintings, sculptures, models as being of whatever objects they represent. The 'possible drawings' which someone who is able to draw can draw do not form a finite surveyable set (or an infinite one). No more so do the 'possible drawings' which someone who can understand drawings can understand. Does it not follow that to characterize what it is that someone who can draw, or someone who can understand drawings, can do requires a recursive characterization of the generation of drawings in terms of their constituent (lines ( ?), shadings ( ?) , etc.) or of their 'meanings' from the 'meanings' of their constituents? Such open-ended abilities are readily characterized, if need be, by examples of their exercises and a similarity rider. Someone who can understand or recognize drawings or paintings can tell you what a Hobbema landscape represents, what a Van der Velde seascape is a painting of, what a Steinberg cartoon wittily portrays or what an Escher etching ambiguously and contradictorily represents, and so on. Such characterizations of the relevant abilities are not defective or inadequate. They do not leave us ignorant of what it is that painters, sculptors, musicians can do, nor of what it is that viewers and hearers who under­ stand what they produce can understand. 73 The open-endedness of the ability to understand a language is held to be captured in the claim that a language is an infinite totality of sentences. But this thesis is trebly misleading. First, the only clear model of a set with an infinite number of members is a set specified by citing a formula which recursively generates its members (e.g. the definition of the natural numbers as the set whose members are generated from the number 0 by iterations of the operation of adding one (the successor operation) ) . Hence, the intelligibility o f treating mastery o f English as represented by an infinite set of pairings of sentences with their truth-conditions pre­ supposes that there is available a recursive mechanism for generating these pairings. But that begs the central question. The need for a theory of

364

365

73 One should not, however, exaggerate. Understanding paintings or engravings, just like understanding sentences, is a matter of degree. One may understand that Diirer's 'Melancholia I' represents a winged figure deeply sunk in thought, surrounded by geometric and craft instruments and accompanied by a putto, without understanding the incono­ graphical significance of the engraving. But equally one might understand Hamlet's 'Get thee to a nunnery' speech without grasping its cruel innuendo.

366

The Generative Theory of Understanding

meaning turns out to be a product of a dubious description of what a language is. Secondly, the alleged 'infinity' of English blurs a crucial distinction. The argument for holding English to consist of infinitely many type-sentences is simply that, given any listing of well-formed meaningful sentences, we could always add one more which is not already on the list. But this does not establish that the meaningful sentences of English form a totality whose members can be correlated one-to-one with the natural numbers. That requires a law for generating an English sentence from each natural number and vice versa (parallel to the operation of multiplication by two which correlates the integers one-to-one with the even integers). In absence of such a law, the possibility of adding yet one more well-formed English sentence to any list of grammatical English sentences shows merely that there is no such thing as the totality (or number) of well-formed sentences of English (just as, though for different reasons, there is no totality of small integers or deeds that a person has performed in his lifetime). Thirdly, even if a mathematician and a linguist together produced a law which effectively correlated each natural number with a grammatical English sentence and vice versa, it would not follow that there is even a possibility of constructing a law which pairs each English sentence with its sense (truth-conditions). That presupposes that the distinction between sense and nonsense applies to each type-sentence independently of the circumstances in which it is uttered (d. above, pp. 2 1 8-28). Once this error is rectified, the idea of characterizing under­ standing in terms of an infinite set of type-acts (of pairing sentences with their senses) becomes ludicrous; for then somebody might ex hypothesi manifest his understanding of a nonsensical utterance by making sense of it! Hitherto we have concentrated on the motivation for the minimalist's making a comparison between mastery of language and knowing how to multiply. This rests on deep confusions. But so too does the analogy itself. It presupposes that a meaningful sentence of English resembles a multipli­ cation problem. But where is the analogy? A sentence is not a problem calling for a solution. Of course, one can pose the problem of what a sentence means (how it is to be understood) ; then giving an explanation of what it means is an answer to this question as well as a criterion for understanding the sentence. Yet uttering a sentence is not to ask what it means, and understanding what is said is not typically manifested in explaining what has been said by another, but rather in responding appro­ priately to it, either in speech or in non-verbal behaviour. Only if a sentence is misconceived as encoding a message is there scope for constru-

, fJi '

l'

I� ,� I'

N

j

J }, ,!,

The Generative Theory of Understanding

367

ing understanding speech on the model of solving multiplication problems. The other half of the analogy is equally strained. What is the counterpart of producing the correct answer to a multiplication sum? Presumably it must be 'assigning truth-conditions to a sentence' ? But, as already noted, this phrase is opaque. It does not designate any act, mental or overt, for which there are any known criteria, and it is doubtful whether any act could constitute an explication of understanding a sentence (since this is an ability). Understanding a sentence comprises a variety of abilities, e.g. the abilities to use it correctly in appropriate circumstances, evaluate the correctness or appropriateness of what is expressed by its use in different contexts, respond appropriately to its use by others, paraphrase it, and explain it in different ways if misunderstandings arise. Many theorists would claim that understanding a sentence is a matter of knowing its meaning and then they might compare knowing what a sentence means with knowing the answer to a multiplication problem. But this argument achieves nothing. For 'knowing the meaning of a sentence', like 'knowing the time', 'knowing the way the wind blows', 'knowing the value of Diirer prints', 'knowing the correct way to behave', designates not an act, but an ability to answer a question. Knowing the meaning of a sentence is knowing what the sentence means, just as knowing the time is knowing what time it is, and knowing the way the wind blows is knowing which way it blows, etc. And to know such things is to be able to answer the questions, here stated in oratio obliqua. If 'assigning a sentence its truth­ conditions' is to explain what it is to know the meaning of a sentence, then it too must designate an ability, not an act. This, however, destroys the analogy with the case of multiplication, where mastery of multiplication is manifested by the acts of producing answers to sums. The analogy would not be restored by substituting 'knowing the correct answer' for 'produc­ ing the correct answer', for that would undermine the idea that the recursion formula for multiplication characterizes what the competent arithmetician can do (since knowing the correct answer is not an act, but an ability too). There is no such thing as an act of assigning truth­ conditions to a sentence which stands to mastery of a language as giving an answer to a multiplication problem stands to mastery of elementary arithmetic. The analogy is altogether lame. The endlessly variegated exercises of the mastery of a language are not characterizable by a 'theory of meaning for a natural language'. Lack of such a theory does not imply inability to characterize what it is to under­ stand a language, nor does it suggest that we know, but cannot say, what such an understanding is, let alone that we do not even know what it is to

fl �.

368

The Generative Theory of Understanding

understand a language, even though we all understand our native tongue. There is here a resemblance between 'understands' and other (by no means homogeneous) cognitive verbs like 'knows', 'believes', or 'remembers'. Remembering, like understanding, does not connote an act or activity, even though one can try to remember, as one can try to understand, succeed in remembering, as one can succeed in understanding, suddenly remember, as one can suddenly understand. The relation between 'remember' and 'can remember' is as slippery as that between 'understand' and 'can understand'. One may fail to remember, as one may fail to understand, misremember as one may misunderstand. Remembering, like understanding, is akin to an ability. For a person who can remember something or other can do numerous things which someone who has forgotten it cannot. He can give correct answers to questions, correct other people's erroneous claims, make true statements on the matter, engage in various activities such as depicting, modelling, re-enacting, or mimicking what it is that he remembers. Nevertheless, it would be wholly misguided to search for an answer to the question 'What does remembering consist in ?' There is no act or activity, either physical and overt (e.g. behaving as if such-and-such were true) or mental and covert (e.g. having a vivacious or familiar image) which is or constitutes the remembering. So too the question 'What does understanding a language consist in?' is confused. To produce a theory of meaning for a natural language in answer to it is to match a nonsensical reply to a nonsensical question. The minimalist enterprise does not succeed in establishing a theory of meaning according to which the meanings of sentences are generated out of their constituents and their structures. Rather, it begs the question of whether there is any such thing, and it ignores arguments that show the idea of such a theory to be incoherent. Like the full-blown endeavour, it founders on fundamental misunderstandings about human under­ standing.

CHAPTER 1 0

Turning Full Circle

1

Kernels of truth?

We have presented a sweeping indictment of modem theories of meaning. Under the guise of rigorous theorizing, philosophers propound as funda­ mental truths nonsensical answers to nonsensical questions. Theoretical linguists too incur the same harsh verdict despite their pretensions to develop an empirical science of language. Our investigations have turned up many absurdities, and further excavations have revealed deeper levels of nonsense beneath the surface strata. It may seem as if our purpose were to show that theories of meaning consisted of nonsense atop nonsense atop nonsense . . . - that they were nonsense all the way down. Against this extreme claim many men of goodwill and sound under­ standing will protest. 'Surely,' they will insist, 'there must be something in all of this work. Confusions, inconsistencies, absurdities, unclarities and exaggerations there may be, but these do not warrant a total rejection of the entire content of every theory of meaning. It must be conceded that there is some connection between sentence-forms and speech-acts performed by utterances, some relation between meaning and truth, and some link between syntactical rules and understanding sentences.. Hence truth-conditional semantics must be acknowledged to rest on unshakeable foundations, even if the superstructure is unstable. Moreover, directing philosophical attention to questions about language is not altogether misguided. For it must be granted that this has yielded much illumination: e.g. it has clarified the nature of valid inference and of logical laws, and it has facilitated the clarification of many psychological concepts. Though the products of the investigation of language may fall short of what enthusiasts claim, they are far from negligible. Would it not be absurd to contend that modem theories of meaning add not one jot or tittle to our understanding?' This objection has great persuasive power. One has a strong inclination

Turning Full Circle

Turning Full Circle

to accept it at once. How could so many well-intentioned and intelligent persons have accomplished nothing at all? Who can reasonably put him­ self in complete opposition to the wisdom of his age? Only the arrogant (or the insane) would fail to accept a truce on these minimal terms. But perhaps this inclination to make peace should be resisted for the moment while the strength of the objection is more carefully weighed. The argument must be that modern theories of meaning have some redeeming features, i.e. that something can be salvaged from them. For were it admitted that only those statements which were antecedent to any theory­ building could be defended, it would also be acknowledged that the theories themselves had achieved nothing of value. A counter-argument is easily developed along these lines: the platitudes cited as the foundations for truth-conditional semantics are typically misrepresented when cast in this role, and when properly understood they owe nothing to any theory of meamng. Consider for example, the connection between sentence-forms and speech-acts. Educated speakers of English do single out a class of sentences termed 'interrogative sentences' or 'questions', and another called 'im­ peratives'. It is platitudinous that interrogative sentences are standardly used to ask questions and that imperatives are typically used to issue orders or to make requests. But such truisms relating syntactic forms of sentences to speech-acts are altogether independent of the thesis that every identification of a speech-act rests on an inference from the form of an uttered sentence to the specification of the speech-act performed. It is far from platitudinous to claim that every intelligible language must provide means of recognizing what use a speaker makes of an arbitrary sentence. But such psychological theses are the stuff of theories of force in philosophy and in theoretical linguistics, and these claims are wrongly taken to be the proper 'scientific explanations' of the prescientific rough­ and-ready generalization that sentence-forms are standardly connected with speech-acts. What is distinctive of theories of meaning are not truisms, and the genuine truisms owe nothing at all to these theories. Similar conclusions hold in respect of other cited platitudes. Consider, e.g., the alleged connection between meaning and truth. If meaning is a feature of sentences and truth a property of statements, and if the notion of a truth-condition is not intelligible, then any possibility of a simple connec­ tion is precluded. A claim that substituting one word for another with a different meaning will typically alter the truth-conditions of a sentence will boil down to the truism that what is asserted in uttering a sentence will typically depend on what words are uttered, hence that changing words

may alter what is said (e.g. transforming a true statement into a false one). A parallel argument connects the syntactic structure of a sentence with what is said in uttering it. But these truisms are quite independent of any thesis distinctive of truth-conditional semantics. Such counter-arguments suggest that nothing can be salvaged from modern theories of meaning. We shall strengthen this case by a synopsis of our earlier criticisms. But first, lest this conclusion be rejected out of hand, we note that the same verdict must be returned in respect of many theories that have exercised the ingenuity and excited the passions of philosophers. For centuries debate raged about universals, as realists and nominalists touted their rival panaceas. Other generations of philosophers warred over the reality of relations and the universal applicability of subject/ predicate decomposition of judgments. Yet others laboured to attack or defend the power of pure reason alone to attain knowledge. And so on. How much can philosophers now salvage from these great controversies? It is striking that most of the arguments no longer make sense, at least as they were then formulated. Important presuppositions are now thought to be incoherent or indefensible, and many crucial concepts are held to be confused. The debates lapse, and great philosophical theories are put away on high shelves and left to gather dust. Truth-conditional semantics would enjoy the solace of having the most distinguished of companions if it were overtaken by a similar fate. No other dead philosophical theory can boast of substantially greater achievements.

370

2

,I

;!i

371

Discredited methodology

One main ground for a wholesale condemnation of modern theories of meaning, whether in philosophy or in linguistics, is a pair of fundamental misconceptions about methodology. Both have been speIt out in the course of earlier criticisms. But their cardinal importance merits their inclusion in the recapitulation of our reasoning. The first of these misconceptions concerns the problems addressed by modern theories of meaning. We have argued that the basic problems are all bogus. Theorists seek answers to a range of distinctive questions: How does one recognize the speech-act performed by a given utterance? How are explanations of word-meanings to be extended to non-declarative sentences ? How is identity and difference of meaning determined? How is the meaning of a molecular sentence generated out of the meanings of its parts? How does one understand sentences never before encountered?

Turning Full Circle

Turning Full Circle

And so on. Each of these questions (at least as interpreted by theorists) makes no sense. Some of them arise out of conceptual confusions (e.g. about recognition and understanding). Others of them grow out of the basic presuppositions of truth-conditional semantics (e.g. that explana­ tions of words are tailored for occurrences in declarative sentences). In the first case, the 'problems' would completely disappear once the misunder­ stood concepts were clarified. In the second, far from the 'theory' provid­ ing the answer to an independent problem worthy of investigation, the 'problem' is parasitic on the 'solution' ; hence, were the 'theory' to be put aside, nothing would remain to be explained. Truth-conditional semantics suffers from a dire disease: it is at a loss to find any genuine problems. The theorist is equipped with some exciting machinery (e.g. the predicate calculus, or a theory of truth for a formalized language); he considers it to deserve wider application (because of the intrinsic beauty of its design) ; b u t h e lacks any raw material which might b e fed into it to some purpose. Casting around, he is apt to latch on to some range of pseudo-problems or else to retreat to some hopelessly vague question (e.g. 'How does language work?'). Even if the consequent 'theories' made sense, they would explain nothing whatever. Would it not be perverse to ascribe intellectual value to statements which cannot be displayed as answers to any intelligible questions? The second misconception concerns the nature of the explanations offered by modern theories of meaning. We have argued that such theories embody radical confusions about the concept of an explanation. They model their explanations on the hypothetico-deductive theories of ad­ vanced physics, proclaiming the virtues of bold postulation of theoretical entities and of unification of hypotheses under covering generalizations. But this concept of explanation is inapplicable to the clarification of normative practices; there a different concept of explanation is applied one in which a rule invoked to explain an agent's (or a group's) actions must itself figure explicitly in the intentions of the agent and in his normative behaviour (e.g. teaching others, correcting himself or others, and justifying actions) . Since speaking a language is itself a normative practice (as many theorists insist), explanations of meanings must take the form of rules which are acknowledged constituents in the practice; such explanations must have normative functions (as cited standards of correct­ ness), and they must be rules actually followed by speakers of the language. Consequently, any 'explanations of meaning' which involve postulating theoretical entities or discovering radically new generaliza­ tions must be excluded a priori. There is no such thing as a mere hypothesis

which constitutes a genuine explanation of the meaning of an expression. Rules, unlike the forces that figure in explanations in physics, cannot 'act at a distance'. In truth-conditional semantics, two distinct concepts of explanation are crossed, and the offspring are 'explanations of the mean­ ings' of particular words which are not explanations of meaning at all. Is there any greater merit to be found in statements delineating the physics of the rules implicated in speaking a language than in statements describing the normative practices of Newtonian point-masses? This pair of methodological misconceptions deprives truth-conditional semantics of any interest. The principles of such a 'theory' serve no serious purpose. They have no place in promoting an understanding of the con­ cepts of meaning, understanding, explaining meaning, language-mastery, symbolism, etc., and they contribute nothing to the clarification of any of the multitude of concepts grasped by competent speakers of natural languages. The point is not that there is no scope for deepened understand­ ing of concepts employed in describing languages and intelligent speech; nor is it that there is no room for various degrees of misunderstanding or lack of understanding of expressions in circulation among competent speakers. Clarification of concepts is always possible and sometimes necessary. But a misguided methodological orientation guarantees that the meanings of expressions, which are aspects of a normative practice of using symbols, hang out of reach of any explanation generated out of a theory of meaning. In the quest for understanding, the cultivation of truth-conditional semantics is a pointless and time-consuming detour.

372

3

g

373

A mythology of symbolism

The second main round for a wholesale condemnation of modem theories of meaning is a host of interlocking and proliferating conceptual confusions. Many have been exposed already. We have argued that rules are conflated with regularities, that understanding is misconceived as a mental process, that explanations of meaning are mistaken for applica­ tions of expressions, etc. We have revealed that such notions as sentence, mood, sentence-structure, truth, synonymy, context-dependence, and nonsense are grossly distorted. And we have indicated that such semi­ technical expressions as 'force', 'truth-condition', 'deep structure', and 'tacit knowledge' are incoherent - the surface manifestations of vain attempts to compensate for underlying confusions. Misconceptions ex­ hibit exponential growth like virulent bacteria. Each one produces many

374

Turning Full Circle

Turning Full Circle

more until the sound tissue of internal relations among concepts gives way under the onslaught. According to our diagnosis, truth-conditional sem­ antics is a vast network of interrelated misconceptions. If these were unravelled one by one, the biological niche occupied by this 'theory' would cease to exist. It would be tedious and pointless to rehearse, or even to recapitulate, the web of earlier arguments demonstrating the conceptual confusions integral to truth-conditional semantics. But we might profitably indicate a number of important misconceptions not previously highlighted. Further attention to these might be a prophylactic against contamination by ones already discussed. Theories of meaning take rise from the preconception that a human language is a system. This faith is deeply entrenched. So much so that theorists treat it as a brute fact that languages are systems. Nobody seems shocked by the suggestion that 'a language is definable in terms of a set of rules . . . which combine with each other to form a system - a grammar which gives us an explicit and exhaustiv� description of every sentence which goes to make up a language. ' 1 Nobody finds it ridiculous to elabo­ rate 'a theory of meaning which represents mastery of a language as the knowledge . . . of deductively connected propositions'.2 The path to these ideas is made smooth by the initial conviction that a language is a system. But the transition is none the less extraordinary. How is the term 'system' to be understood? What justifies its application to a natural language ? And, more importantly, what are the implications of characterizing English as a system? We speak of systems of musical notation, systems of weights and measures, systems of transport, systems of criminal law, judicial systems, systems of education, systems of political checks and balances, electoral systems, etc. But some systems are, as it were, more systematic than others. Who would not scoff at the proposal to exhibit the network of trunk roads in Britain as the output of a calculus imposing a minimizing constraint on satisfaction of needs to travel between different points? One is predisposed to view the road system differently, as the more or less haphazard product of a host of piecemeal improvements (and natural

disasters !) spread over two millennia. Would it not be at least as plausible to consider the English language to be a similar system? Has it not also evolved over a long period by gradual accretion and loss of vocabulary, phrases, and grammatical constructions? It is certainly open to us to look at a natural language as a loosely integrated normative practice, a motley of rules on a par with the common law. Once this possibility is acknowl­ edged, our viewing English under the guise of an axiomatic calculus must be considered a matter of choice. The crucial questions then become what purposes are served by making this decision, and what intellectual costs are incurred. And these questions would lead back to a careful scrutiny of the concept of a system. This seems an urgent task to forestall the spinning of fairy-tales about natural languages. A second array of unclarities surrounds the applications of the term 'language'. By ignoring important distinctions and disregarding certain internal relations, theorists of meaning fall into a myriad of misconcep­ tions. One range of confusions arises out of insouciance about the distinc­ tion between a sortal noun ('a language', 'languages') and an abstract mass-noun ('language' simpliciter). We speak of English, French, Russian, etc. as particular languages, and we may contrast the colour-vocabulary in our language with that of another. Each language has its own grammar, and hence its syntactic structures may differ from those of other languages. Similarly, each language has a distinctive vocabulary, and this is linked with a distinctive practice of explaining the meanings of these words. Linguists and philosophers are aware of all these platitudes, but they readily lose sight of them in the desire to achieve greater gen�raijty or profundity. Linguists slip into speaking of 'language-learning abili tY ' and 'the depth-structure of language' rather than sticking with 'English­ learning ability' or 'the depth-structure of Russian'. And philosophers investigate 'the logical structure of language' (not of English or French), 'the isomorphism between language and the world', or 'how language connects with experience'. This practice is well-entrenched, yet none the less dubious. Any question about the generality of a theorist's observations is openly begged, and a specious air of profundity is produced by verbal sleight-of-hand. Parallel questions framed with abstract mass-nouns are apt to appear silly; e.g. 'What is the relation between law and human action?', 'How does money connect with goods and services ?', 'What is the essential structure of literature ?', or 'Is music-learning ability acquired or innate ?'. And equally ridiculous are the results of retreating from the mass-noun 'language'; who would wish to explore the logical structure of French, the possibility of isomorphism between Algonquian and the

I N. Smith and D. Wilson, Mcdern Linguistics: the Results of Chomsky'S Revolution (Penguin, Harmondsworth, 1 979), pp. 13f.; d. N. Chomsky, Syntactic Structures (Mouton, The Hague, 1 957), p. 18. 2 M. A . E . Dummett, 'What i s a Theory o f Meaning?', in Mind and Language, ed. S. Guttenplan (Clarendon Press, Oxford, 1975), p. 1 12, d. 'What is a Theory of Meaning? II', in Truth and Meaning, ed. G. Evans and J. McDowell, (Clarendon Press, Oxford, 1976), p. 70.

3 75

376

Turning Full Circle

Turning Full Circle

world, or how Hopi connects with experience? Is there any cogent reason for supposing all languages to have the same 'logical structure' or to 'connect with experience' in the same way? Perhaps the sole foundation for philosophy of language [sic !] and theoretical linguistics is a misuse of 'language' . Another range of confusions concerns the boundaries of languages. 3 Theorists carry on as if what belonged to a given language were a clear datum, beyond doubt or controversy. They think of a language as consist­ ing solely of verbal symbols, words both spoken and written; or they characterize a language as an infinite set of meaningful sentences. Both syntax and semantics study only such symbols; everything else, whether gestures, samples, non-verbal voiced sounds, social conventions, etc. is relegated to the separate science of pragmatics. Mastery of a language is held to be limited to the ability to assign the correct meaning to every sentence of the language. All of these contentions manifest a distorted conception of speaking a language, and each of them propagates further confusions. Gestures, samples, tones of voice, etc. often play crucial roles in communication, and they often qualify as symbols (as theorists ac­ knowledge in speaking of gesture-languages). Many of them are both specific to particular languages and capable of explanations parallel to explanations of the meanings of words. Is there any cogent reason for putting them beyond the pale of particular languages apart from their manifestly failing to fit into any version of a formal calculus? What purpose is served by excluding French gestures from the investigation of the French language? The consequences of this excessive narrowing of the concept of a language are clearly disastrous. One is a ridiculous separation of the question of whether an utterance makes sense from the question whether it is intelligible; e.g. there would be no ground, on this view, for calling into question my understanding of English if I say 'Her hair is this i colour' while pointing at a man and producing a tone with a tuning fork. Another consequence is a distorted notion of mastery of a language. Somebody who has perfect mastery of English must be capable of carrying on conversations, making and responding to requests, asking and answer­ ing questions, etc. But these aspects of speaking English are interwoven with a vast range of other activities, and they interlock with many kinds of skill and knowledge. Intelligent speech manifests general savoir faire as

well as familiarity with social customs or practices and specific knowledge about architecture, engineering, literature, gardening, games, etc. The ability to speak English merges into such other abilities as social skills, memory, motor control, and the capacity to articulate knowledge clearly. It is a mistake to consider that mastery of a language is an ability which is sharply circumscribed and properly described independently of other abilities. In particular, it is wrong to isolate understanding from behaviour by denying that somebody's saying something unintelligible or his reacting inappropriately to an utterance are criteria for his not correctly under­ standing what is said. The corollary of misdrawing the boundaries of a language is a set of confusions about the concepts of understanding and of mastery of a language. The misconception about the boundaries of a language is but one manifestation of a confused conception of symbols. Theorists tend to think that words and sentences are intrinsically symbolic, as if these were the only genuine symbols and other things (gestures, samples, diagrams) were symbols merely by analogy. The truth, however, is that classification of something as a symbol depends on how it is used. If it is to be seen within a particular community as conveying a message or expressing something, then it is a symbol (at this time for these persons) . Nothing has an inalienable status as a symboli even words, sentences, or monograms can be used for mere decoration or to compose concrete music (e.g. a word­ fugue). Conversely, nothing perceptible is excluded from functioning as a symbol; it may be incorporated as an element in a code, a signal, or a sample. Something is a symbol if it is to be seen as a symbol. Consequently, it is a mistake to accord words and sentences some privileged status in the clarification of the concepts of a: symbol and of communication by symbols. Theorists make other errors as well. One is to analyse sentences into depth-structures whose elements have no realizations at all. As if there could be such a thing as a symbol which is in principle not perceptible! Both this error and the opposite are also widespread in the discussion of the type/token distinction. A written token-word is identified as a physical object, and hence no such single token can intelligibly be said to occur in two places at the same time. 4 1t must be a symbol which cannot in principle be replicated. On the other hand, we say that the same word occurs twice on a single page of a book; hence, what occurs twice cannot be a physical object, but must be an abstract pattern or design of which two separate realizations occur on the page. This type-symbol, being an abstract object,

-' One (not further discussed here) is a distortion of the principles for individuating languages that springs from philosophers' claims that a 'metalanguage' must be a distinct language from its 'object language'.

4 Spoken tokens raise other problems, equally acute.

377

Turning Full Circle

Turning Full Circle

is veiled from the senses by its tokens. As if there were such things as symbols which cannot be reproduced and symbols which cannot be perceived ! These muddles are symptomatic of deep misconceptions about symbols. They would be remedied only by a thorough-going clarification of how concepts of perceptible objects (and events) are related to the concept of a symbol. To this catalogue of confusions about very general and basic concepts must be added a distorted conception of understanding. Theorists tend to treat understanding as a form of information-processing, and this generic activity is further likened to machine-translation. Theories of meaning present 'models' or 'representations' of understanding in the form of mechanical derivations within formal calculi. This outlook produces many absurdities. One obvious consequence is that understanding a sen­ tence is analysed as translation from one symbolism into another (perhaps into 'the language of thought' !) ; but this generates a vicious infinite regress. Another consequence is to demote various criteria of understand­ ing to the status of mere inductive evidence. In particular, how someone reacts to utterances depends on factors other than his ability to 'decode' what the speaker has said, and hence it appears to have no direct bearing on his understanding; but this conclusion misrepresents the concept of understanding in an important respect. Yet another consequence is to impoverish this concept by neglecting vital internal relations with other concepts. One of these is the relation of understanding to explanations of meaning. Another is its relation to mastery of a practice and indirectly to imagination. This can be illustrated by analogy. Somebody who has mastered chess may see his opponent's pawn-move as a threat to his distant king, whereas a neophyte could not; or somebody familiar with solid shapes and the physical possibilities of their orientation might see a diagram as a triangle which has fallen on to its side, whereas a very small child or a cat could not. In such cases, mastery of certain techniques and certain powers of imagination are crucial to the 'interpretation' of what is perceived. The same conceptual links are obviously vital to understanding human speech and writing. A modest degree of imagination is required to see the remark 'It's getting rather late' as a request to leave, and a far more complex conceptual background is needed to see this paragraph as part of a criticism of truth-conditional semantics. Construing understanding on the model of derivations within a calculus screens from theorists' view many significant aspects of the concept of understanding. Unless these exploratory remarks about the concepts of a system, language, a language, a symbol, and understanding are utterly misguided,

modern theories of meaning should not be conceded to have gone off the rails somewhere down the line. Rather, they never succeeded in putting themselves on the rails at all. Or, to change the metaphor, they rest on confusions that extend to rock-bottom. Indeed, such theories are simply the products of conceptual confusions. For various reasons, we are all prone to misinterpret certain concepts that we use without any hitches in ordinary discourse. Philosophers, being peculiarly sensitive to these influ­ ences and brazen in developing their consequences, give accounts of these concepts which are at first subtly askew and which become more and more outlandish. Then complex theories are adumbrated to compensate for nonsense. Unclarities about the concepts of words, sentences, languages, understanding symbols, explanations, etc., do not issue in the clarification of these concepts through careful exploration of how problem-causing words are explained, understood, and applied. Instead, these unclarities about concepts are misinterpreted as mysteries about phenomena and hence serve as the impetus to searching for explanatory empirical theories. Confusions about the concept of understanding are projected on to the world by considering understanding to be a queer and very complex mental process whose nature must be revealed by psychological investiga­ tions. Mysteries about depth-structures, tacit knowledge, truth­ conditions, logical forms, and so forth are parallel manifestations of conceptual confusions. So too is racking one's brains over the intractable question of revealing how language works. No explanatory theory can alleviate the intellectual discomforts caused by the difficulty of making perspicious how concepts interlock. Anybody satisfied by truth­ conditional semantics must be twice duped and doubly confused.

378

I ·

4

379

Forms of representation

Conceptual confusions may be diagnosed as the nutrient in which modern theories of meaning prosper. But they do not alone account for the particular directions in which these theories grow. This depends on vari­ ous features of the intellectual climate of the present century. Trying to catalogue these influences would be an enormous undertaking, and there would be no agreed standards for deciding when the list was complete. We have indicated en passant a number of influences that seem important. Among these are the status of Euclidean geometry as the model of a perspicuously articulated a priori science, the growth of computer tech­ nology and the investigation of 'artificial intelligence', the rise of a be-

Turning Full Circle

Turning Full Circle

haviourist psychology based on the notion of conditioned responses to stimuli, and the evolution of new forms of mathematics. But above all we have stressed the overarching importance of two influences on theories of meaning: the 'Augustinian picture of language' and Frege's invention of the predicate calculus. These jointly inspired the creation in the 1930s of the science of logical semantics, and its leading ideas have informed subsequent elaboration of theories of meaning in both philosophy and theoretical linguistics. The shape of things even now unfolding is foreshadowed in an age-old picture of the essential role of words and sentences and in a sophisticated mathematical symbolism for representing inferences among complex generalizations. Recognizing these roots is vital for understanding theories of meaning - and their allure. We have presented this synopsis of the development of twentieth century reflections about language in various historical sketches, and we have supported it with detailed analysis of modern ideas. This account informed our original selection of the Tractatus as the pivotal influence on the growth of theories of meaning. For there Wittgenstein manifestly put Frege's new logic to use in vindicating a particularly stark version of the Augustinian picture of language. The Tractatus recapitulates in micro­ cosm the phylogeny of truth-conditional semantics. This point is particu­ larly worth noting since the sins of the Tractatus have been visited on its intellectual offspring. In order to explore this baleful influence, we must once more highlight certain main features of the Tractatus. This is best accomplished by contrast. Let us recall the philosophical foundations of Frege's invention of the predicate calculus. He first proposed his concept-script and his formalization of inference for the specific purpose of checking the cogency of proofs in mathematics; his aim was explicit and limited. He took the subject-matter of logic (i.e. what he analysed into function and argument) to be propositions, judgments, or 'judgeable-contents', not declarative sentences expressing them; he followed traditional wisdom in stressing that the grammatical forms of sentences are often misleading guides to the logical forms of judgments. He emphasized and exploited the possibilities for analysing a single judgeable-content in different ways into functions and arguments of different levels. Finally, he argued that an expression contributes to what a sentence expresses only if it stands for some object or function; hence, in particular, he held logical constants to name certain functions and number-words to name certain objects. It is noteworthy that some of these ideas underwent changes and diminished in importance in the course of Frege's career. In particular, he came to view predicates of

declarative sentences a s themselves names for functions ('concepts'); ac­ cordingly he adopted the idea of a relatively close correspondence of the grammatical forms of sentences with the logical forms of the propositions expressed by them, and ultimately he was seduced by a vision of an isomorphism between sentences and what they express to the detriment of his doctrine about possibilities for alternative analyses of judgments. Yet the origins of Frege's invention of the predicate calculus lay in a clearly articulated set of doctrines uncontaminated by these later intrusions. The Tractatus was the culmination of a drift away from these doctrines in the direction already adumbrated in Frege's own thinking. First, logic evolved into part of a general theory of symbolism; its subject-matter was no longer thought to be abstract entities (propositions or thoughts) as opposed to declarative sentences, but rather it came to be viewed as the investigation of what sentences express through considering the features of these sentences themselves. Secondly, the flexibility of function­ theoretic categories facilitated the logical analysis of aspects of expressions traditionally accorded no explicit attention; e.g. the definite article could be exhibited as a second-level function mapping a concept on to an object (Frege) or as a complex quantifier (Russell) . Thirdly, logical analysis of sentences into function and argument became infused with the notion of conceptual analysis prominent in classical empiricism. 'Molecules' are step-wise broken down into simpler constituents, ultimately into 'atoms' ; the conviction blossomed that every significant sentence has a unique ultimate analysis, and apparently different analyses must represent differ­ ent stages on this route to the final analysis. Fourthly, Russell's theory of definite descriptions entrenched a modified conception of analysis: an 'incomplete symbol' is an expression which has no (independent) meaning but which none the less contributes in a systematic way to the significance of any sentence in which it occurs. Accordingly, it became possible to analyse an expression, not by explaining what it means (stands for), but by explaining it away in favour of other expressions which do have independ­ ent meanings. Finally, the predicate calculus ceased to be regarded as an instrument employed for particular purposes, but instead was held to embrace the logical structures of all fully analysed sentences. Logical analysis reveals every judgment really to be built up out of objects and concepts by means of quantifiers and propositional connectives. The Tractatus exemplified all of these developments in a strikingly spartan form. Every proposition was claimed to be a truth-function of elementary propositions, each of which consisted entirely of expressions standing for entities. With the exception of logical connectives, category names and

380

381

Turning Full Circle

Turning Full Circle

numerals, every expression conformed to the Augustinian picture, and the exceptional expressions were explained in terms of their contributions to the senses of the sentences in which they may occur or else placed on the index. Analysis was held to show that sentences, thoughts, and facts were isomorphic, each with the other. Wittgenstein later came to the realization that much of the thinking in the Tractatus manifested a disease of the intellect. He gave careful scrutiny to the nature and sources of the underlying philosophical illusions, and he catalogued his diagnostic insights in the Philosophical Investigations. These are worth pondering. The fundamental error is a fascination with or captivation by a form of representation. For a variety of reasons we have strong inclinations towards depicting matters in certain ways. It seems deeply satisfying to view words as names and to present inferences in the guise of transformations within a mathematical calculus - so satisfying that we are willing to turn a blind eye to hosts of obvious and powerful objections. Tensions persist. But they are resolved in characteristic ways. One is to offer a philosophical theory as an ideal to which language or thought only approximates (e.g. to follow Russell in considering the Tractatus to outline the structure of an 'ideal language'). The other is to present the theory as an analysis of what it represents, as delineating the reality masked beneath misleading appearances (e.g. to hold the Tractatus to lay down the hidden structure of any possible language capable of expressing truths about experience) . Each of these manoeuvres is con­ fused. The first overlooks the truism that what is ideal for one purpose may be inappropriate for another, or it presupposes the absurdity that there are purpose-independent standards for what is ideal. The second falls into a related misconception, as if there were an absolute standard of what counts as an analysis (and corresponding absolute standards of simplicity and complexity) . The second also exemplifies dogmatism. For no good reason it is insisted that things must be thus-and-so (e.g. that every sound inference must exhibit a pattern validated by the predicate calculus, or that every meaningful expression must stand for some object or function). This unshakeable faith justifies confidence that it will one day be discovered exactly how these requirements are met, even if we now have no inkling of how this might be done. The mistake here is to 'predicate of the thing what lies in the method of representing it'. 5 The primary symptom is grotesque misuse of expressions, which can be made clear by careful comparisons between philosophers' uses of crucial expressions and the down to earth

uses o f these expressions in everyday discourse (e.g. 'sentence', 'object', 'fact', and 'symbol'). As philosophers, 'impressed by the possibility of a comparison, we think we are perceiving a state of affairs of the highest generality. . . . We can avoid ineptness or emptiness in our assertions only by presenting . . . an object of comparison . . . as a measuring-rod, not as a preconceived idea to which reality must correspond'. 6 The danger is projecting a form of representation on to the phenomena represented. This diagnosis of philosophical illusion contains not only a negative view about misuses of forms of representation, but also positive sugges­ tions about their proper use in philosophy. This is important. The corre­ late of Wittgenstein's criticism of the Tractatus is not that the predicate calculus must be jettisoned, or even that we must dismiss as worthless any suggestion that the meaning of a name is the object named or that sentences are pictures of states of affairs. Rather he indicates that the very same ideas which typically mislead philosophers have potentialities to serve as illumi­ nating objects of comparison. We may profitably compare a name to a label attached to an object, a sentence with an architect's plan, or a stretch of scientific discourse with a complex arithmetical calculation, provided that we do not exaggerate the similarities or neglect the differences. It is crucially important to bear constantly in mind the purposes for which the comparison is made, and to be clear about the purposes thwarted as well as those served by the comparison. A well-chosen model may help to make perspicuous certain matters whose significance is not appreciated just as juxtaposing a familiar piece of music with another one may throw some of its features into high relief. Philosophical illusion often arises from the fact that a comparison is illuminating: we are prone to think that any greater understanding must be the consequence of discerning some deep pattern capable of fruitful generalization, and we rush ahead into formulating explanatory theories to satisfy this craving for generality. To counteract these tendencies to error we should remember that comparisons are not primitive hypotheses, that limitations on comparisons can be as illuminat­ ing as exact correspondences, and that confusion is the price of extending comparisons beyond their intrinsic limits. More positively, we should remind ourselves of the purpose-relativity and the limitations of compari­ sons by deliberately exploiting many, not just one, in respect of any matter requiring philosophical clarification. It may be useful to compare naming something with attaching a label to a thing, but equally useful to compare a name with a definite description or a demonstrative, and using a name

5 Wittgenstein, Philosophical Investigations, § 1 04.

6 Ibid., § § 1 04 and 1 3 1 .

3 82

383

384

Turning Full Circle

Turning Full Circle

with summoning somebody or transfixing him with an arrow. Similarly, it may be helpful to compare an expression of generality with a second-level mathematical operation such as differentiation, but equally important to distinguish quantifiers from function-names or to differentiate among the patterns of inferences licensed by various 'universal generalizations'. Objects of comparison are not misleading in themselves. It is misuses of comparisons that are misleading and correspondingly their judicious use which promotes philosophical understanding (a synoptic view or Ubersicht of relations among concepts). Whereas the illusions of the Tractatus stemmed from projecting methods of representation on to generalizations about the phenomena represented, the insights of the Investigations arise from careful selection of objects of comparison with the conscious purpose of spotlighting significant resemblances and differences among concepts. Wittgenstein castigated the Tractatus, and many of his trenchant criti­ cisms are now widely accepted. Most philosophers now ridicule the notion of a logically proper name, and they also repudiate the contention that every significant sentence must be a truth-function of any of its constituent sentences. Such reactions have meant that rejection of the Tractatus has taken place independently of any exploration of the underlying general reasons for its defects, as if all of its errors were matters of comparative detaiL This in turn has facilitated the growth of more sophisticated but closely allied doctrines within truth-conditional semantics. The Tractatus has been reborn, allegedly purged of its initial crudities by progress in formal logic and exploiting the resources of appropriate metalanguages. The acme of formal logic is the working out of a precise semantic concep­ tion of validity for inferences in the predicate calculus. This turns on generalizations over all admissible interpretations of well-formed formulae, and that allows logicians to cut themselves free from any abso­ lute notion of logical form and hence from the metaphysical idea that logical truths mirror a language-independent structure of the world. Logic is no longer fettered to any doctrine of analysis. What has taken the place of the thesis that every significant sentence is a truth-function of elementary sentences is a much more liberal and characteristically second­ order claim. One popular version is a truth-theory for a language: a statement of the truth-conditions for any significant sentence must be derivable within an axiomatic metatheory by appeal to rules of designa­ tion for its constituents together with rules specifying how the designa­ tions of well-formed combinations of constituents depend solely on the designations of their parts. Another popular version builds in a similar metalinguistic theory about the intensions of sentences. Both officially

renounce the pretension that a proper semantic analysis of language be unique because it is an article of modern faith that any number of coherent theories are compatible with any body of 'data'. Semantics and logic are linked in the doctrine that a correct theory of meaning for a language must exhibit logical truths as sentences true under all admissible interpretations of their constituent terms. Whereas the Tractatus considered analysis to be translation into a perspicuous notation, modern theories of meaning rest on metalinguistic generalizations about all assignments of interpretations to expressions which are compatible with judgments about the truth­ values of speakers' statements. This modern wisdom draws on the same sources of inspiration as the Tractatus. It is still obsessed with the predicate calculus; the 'workings of language' are held to be perspicuous in the case of the formal semantics for the predi�ate calculus, and the conscious aim of theorists is to extend this successful model to account for the meanings of all expressions in a language. Attachment to the Augustinian picture is equally clear, since the notion of an interpretation in formal semantics is explained as a correla­ tion of expressions with suitable entities (which thereby correlate sentences with sets of truth-conditions). Although differently expressed, the con­ tinuity of inspiration is evident. So too is the underlying defect · of the Tractatus: modern theories project the method of representation on to what is represented. This generates characteristic dogmatism: every sentence must decompose into a separate force-indicator and sentence-radical, and the truth-conditions of any sentence must be derivable from its con­ stituents and its structure. Some fresh arguments (especially deductions from the possibility of understanding new sentences) and some novel conceptions (especially about data-processing and computer technology) feed this dogmatism. Furthermore, some fresh expressions are given to the dogmatism itself (especially patter about deep structures of sentences and the projection of function-theoretic categories on to the syntax of sen­ tences). But the continuity with the dogmas of the Tractatus stands out. Modern dogmatism is damned for the same general reasons as its pre­ decessor. Its roots are conceptual confusions and infatuation with pre­ conceived pictures. The mystery of the 'creativity of language' expresses the fundamental misapprehension of understanding as a mental process whose mechanism must be brought to light. The insistence that speaking a language is closely analogous to operating a mathematical calculus mani­ fests a misplaced faith in hidden system. Apart from philosophical clari­ fication aimed at bringing light into these dark corners of the modern Zeitgeist, the most effective therapy for modern dogmatism is to press each

385

386

Turning Full Circle

Turning Full Circle

theory for answers to the questions 'What are the purposes it serves?' and 'What are the distortions it introduces?' These are the basic issues. Attain­ ing a proper grasp of problematic concepts is more akin to coming to a full understanding of a genre of music than to apprehending the laws of motion of the planets. The value of an explanation 'cannot here be measured independently of the purpose that it is meant to fulfil, and there are excellent reasons for thinking that none will be absolutely the best (i.e. against the background of any conceivable purpose). Few are useless, but none is omnicompetent. Consequently, it would likewise be wrong to react to modern theories of meaning by repudiating all of their tenets or all of their apparatus as utterly misguided. The predicate calculus has a legiti­ mate role in clarifying the forms of certain inferences, especially for mathematicians who find its function-theoretic concepts perspicuous; but it does not delineate the essence of a language. Similarly, model theory has a respectable, if limited, role as a mathematical investigation of formalized reasoning; but it has no intrinsic philosophical cachet, and it cannot be presumed to illuminate anything whatever about natural languages. It is a platitude that what 'Socrates is wise' expresses is true if the person desig­ nated by 'Socrates' has the property of being wise; but it is mistaken to take the predicate 'is wise' to designate a property or the sentence 'Socrates is wise' to be true, a fortiori to erect a theory of truth on these foundations. And so forth. The proper reaction to philosophical criticism here is to dethrone the dogmatism of modern theories of language, but not to line up all of its deliverances before a firing squad. To the extent that philosophi­ cal illusion arises from projecting forms of representation on to what is represented, the error is attractive precisely because particular compari­ sons may be genuinely illuminating up to a certain point and for certain purposes. The antidote is not an absurd denial that these comparisons have any utility at all, but rather a recognition of points of dissimilarity and of purposes not served. The cure is no less subtle than the disease.

Because philosophy has, as its first if not its only task, the analysis of meaning, and because, the deeper such analysis goes, the more it is dependent upon a correct general account of meaning, a model for what the understanding of an expression consists in, the theory of meaning, which is the search for such a model, is the foundation of all philosophy . 7

5

Philosophy of language

Philosophers bent on constructing theories of meaning for natural languages have arrogated to themselves the phrase 'philosophy of language' to label the product of their activities. This is now considered to be one of the main subdivisions of philosophy, on a par with ethics, metaphysics, formal logic, epistemology, and philosophy of science. Some have made more grandiose claims on its behalf:

.

3 87

.

It is even maintained that this shared faith is the hallmark of modern analytic philosophy: 'the philosophy of language is seen both as that part of the subject which underlies all the rest and as that which it is currently most fruitful to investigate'. 8 Certainly the effort devoted to arguing about the general form of a theory of meaning has waxed in this last decade, while the piecemeal clarification of particular problematic concepts has waned. Philosophy of language has assumed the status of the Young Pretender. The thesis of the centrality of philosophy of language is difficult to support even in theory, and the faith has not been cemented in place by any noteworthy works. If the critical arguments of this book are not totally awry, this failure is hardly suprising. Philosophy of language, as com­ monly conceived, has no coherent subject-matter, and its artefacts are uniformly nonsensical. There is no such thing as a theory of meaning for a language, and hence there is no such thing as a significant contribution to such an enterprise. Any claim by philosophy of language to the throne in philosophy is specious. The title, however, might be transferred to other undertakings which have real content and point. One of these would be the subject earlier called 'philosophical logic'. This attempts to clarify the role of certain kinds of expressions against a background of conceptual problems or perplexities. In the past philosophers have tried to elucidate how various kinds of referring expressions function (e.g. definite descriptions, demon­ stratives, proper names introduced by ostensive definition, or proper names of historical figures) or what kinds of speech-acts are performed by making moral pronouncements (e.g. evaluation, commendation, prescrip­ tion, or prohibition) . Such enquiries are legitimate responses to particular kinds of puzzlement (e.g. the uses of proper names in works of fiction or the occurrence of definite descriptions in counterfactual conditions) . Ex­ planations of how expressions are correctly used may clear away diffi7 M. A. E. Dummett, Frege: Philosophy of Language (Duckworth, London, 1973), p. 669.

M. A. E. Dummett, 'Can Analytical Philosophy be Systematic, and Ought it to Be?', in Truth and Other Enigmas (Duckworth, London, 1 978), p. 441. •

388

Turning Full Circle

cui ties, especially those that beset philosophers in search of generalizations about the employment of expressions of certain forms. Provided that the problems are genuine, there is scope for philosophical investigations to resolve or dissolve them. In so far as a large part of the philosophical technique of so doing consists of detailed descriptions ofthe use of various grammatical categories of expressions, this activity might be termed 'philosophy of language'. It would be one part among others of a general striving after philosophical clarity. Another study that might inherit the title 'philosophy of language' is the clarification of concepts used in describing and analysing languages. This would embrace the investigation of the uses of such terms as 'language', 'a language', 'system', 'rule', 'sentence', 'symbol', 'meaning', 'explanation (of meaning)', 'true', and 'context'. It would also encompass such semi­ technical or technical jargon as 'truth-condition', 'sentence-radical', 'context-dependent', and 'verification-transcendence'. For the present, the primary purpose of this study would no doubt be a general criticism of the misuse and misexplanation of these expressions by theorists of meaning. For, as we have argued at length, these concepts are severally distorted and the internal relations among them neglected or misrepresented. Natural languages are mistakenly treated as formal calculi, and conducting intelli­ gent conversation likened to carrying out formal derivations. The notion of a symbol is perverted, and the relation of meaning with explanations of meaning is slighted. These misconceptions proliferate and nurture each other. Nothing less than a fresh start is needed, a total reorientation of thinking about languages based on a sound understanding of basic con­ cepts. A critical 'philosophy of language' will emphasize that whether something is a symbol depends on how it is used, that the meaning of an expression is what is explained in explaining correctly how it is to be used, that understanding is an ability not a mental process, that the explanation of a word has the status of a rule or standard of correctness, and that whether a sentence formulates a rule depends on how it is employed. The widespread acceptance of truth-conditional semantics shows how far most theorists of meaning are removed from any grasp of these fundamen­ tal aspects of the concepts that they invoke and misuse. And it is important to correct these misunderstandings because their pernicious influence is disseminated through much of contemporary thought, from molecular biology to literary criticism. This book is intended to be a start on this task of philosophical clarification. This critical 'philosophy of language' might be considered a prelude to fresh attempts to construct theories of meaning for natural languages. If

r

j;'I, ;i

Turning Full Circle

389

contemporary work is indeed flawed by conceptual confusions, then might the clarification of basic concepts not permit rebuilding of the envisaged superstructures on solid foundations? If our arguments carry any weight, this hope is vain. Theories of meaning are not merely confused, but also lack any purpose. The clarification of concepts will not conjure up the need for a theory where there was previously no scope for theorizing. On the contrary, it will annihilate the impression that explanations are needed by removing the confusions which are misconstrued as mysteries about phenomena. Theories of meaning will appear ridiculous because they will be seen to address such bogus problems as how somebody recognizes what speech-act is performed by an utterance, how new sentences are under­ stood, or what determines the grammaticality of an utterance. Truth­ conditional semantics will suffer death by a thousand clarifications. It will not be ranged with Newtonian mechanics and Mendelian genetics in the museum of the history of science, but it will be relegated to the basement to moulder away in the company of phrenology and the theory of the humours, the New Way of Ideas and the Theory of Forms. Philosophical criticism should seal its fate. A viable theory of meaning for a natural language cannot rise like a Phoenix if truth-conditional semantics is prop­ erly reduced to ashes.

, , 1'

�:'��

Index

ability, 282-3, 323, 340n, 349-5 1, 354, 362-8 theoretical representation of, 324-6, 3 5 9-68 Allen, ]. P. B., 292 Alston, W. P., 68, 70n ambiguity, 292-3, 3 1 0, 3 3 1 analysis, 45, 3 8 1 principle of, 135-40, 1 9 1 'and', 1 7 1 -4, 1 78-9 anti-realism, 139n, 155-7; see also realism/anti-realism appearance/reality, 14- 15 Aristotle, 1 , 1 7n, 27, 3 1 assertion, 34-6, 56-7 assertion-sign, 55-7, 63, 74, 104-5 Athanasius, 23 Augustinian picture of language, 19, 3 7-8, 4 1 , 146, 238-9, 3 80, 382, 3 85 Austin, J. L., 3, 4, 60-4, 70, 71n avowals, 220 Ayer, A. J., 3

Bolzano, B., 29 Boole, G., 2, 25, 30-3 'but', 1 73

calculus ratiocinator, 23, 32, 36

Carnap, R., 3, 4, 47, 123, 127, 139-43, 145, 147-53, 157-8, 160, 1 64, 1 9 1 , 1 93, 2 1 1n Cartesian method, 1 9 category restrictions, 228, 336-7 central state materialism, 300n Chafe, W. L., 1 8n characteristica universalis, 23-4, 36 Chomsky, N., 8, 10, 67, 73, 1 1 7, 1 69, 247, 256, 276, 280-8, 290-4, 3 0 1 , 306-7, 309n, 3 1 1 - 14, 321, 325-6, 343-7, 354 Cicero, 274 Clark, K., 226 cognitive psychology, 247 'cognize', 247-8, 284-5, 291, 3 12, 340, 343-5 Cohen, L. ]., 23n competence, 8, 67-8, 267, 28 1-3, 3 1 1-12 Bacon, F., 14, 2 1n, 23 completeness proofs, 1 3 1 Baier, K., 8 1n Baker, G. P., 1 3n, 1 9n, 35n, 64n, 142n, componential analysis, 328-32, 336 compositionalism, 213, 2 1 9-20, 3 16, 1 53n, 1 92n, 2 1 7n, 253n, 279n, 33 7-40, 360 320n, 332n, 348n, 349n concept, as function, 34 Bartsch, R., 277n objectivity of; 302-4 Bentham, J., 252 concept-script, 32-3, 36, 55-8, 133 Berkeley, G., 1 6n, 1 9, 43n conceptual change, 163-7 biplanarity, 7, 1 9, 3 7, 268, 276n condition, concept of, 192-6 bipolarity (of proposition), 41 consistency proof, 130- 1 bivalence, 125, 156 context, concept of, 225-6 Bloomfield, L., 8

392

Index

context-dependence, 155, 157, 1 8 7, 1 9 7-205, 2 1 8 contextual principle, 35-6, 1 24-5 conventionalism, 43-4, 305 copula, 27 corpuscularian theory, 14, 1 6 correctness o f semantic theories, 158 creativity of language, 321 -2, 338 Davidson, D., 5, 1 0, 47, 70n, 88n, 98-9, 1 2 1 , 1 22, 1 24, 154n, 157, 1 6 1 , 1 66, 1 68-9, 1 80-2, 1 89, 1 98-9, 229n, 232, 246-7, 3 2 1 -4, 357 deep structure, 8-9; see also depth grammar defeasibility, 78, 1 1 9, 1 92 definition, 1 14, 329 Delgano, G., 23 De Morgan, A., 25, 3 1 depth grammar, 3 1 , 73-5, 89n, 93, 1 08n, 1 13, 288, 309, 3 1 8, 353 De Quincey, T., 83 Descartes, R., 1 4, 1 9, 23, 26-7 descriptions, 39-40, 45, 60, 8 1-7 Russell's theory of, 38, 203n, 233, 381 descriptive content, 6 1 , 63, 72, 78, 80-7, 1 03-4 descriptive fallacy, 60 dictionary definitions, 214- 15, 22 1, 228, 230, 233, 274-5, 33 1 dogmatism (in philosophy), 382, 385 duality, principle of, 1 3 1 Dummett, M. A. E., 4-5, 9- 1 0, 50, 69, 80, 122, 1 25-6, 155, 1 6 1 , 1 66, 1 80- 1, 1 96n, 2 1 1n, 229n, 235, 240n, 241, 247, 276, 321-5, 340n, 347, 357n, 359, 374, 387 elementary propositions, 39-42, 45-6 empiricism, 16, 1 8 entailment, 129 equipollence, principle of, 132-4, 1 70 Erdmann, B., 29 Ernst, P., 1 66 eternal sentences, 1 98, 202-4 etymology, 274

,

Index

Evans, G., 294-5, 340 events, 246 exact, 2 1 7 explanation, 1 83, 372-3 of word-meaning, 1 14, 124, 142, 1 5 1 , 1 74-8, 2 1 On, 214- 1 7, 221 , 230, 232-8, 253, 332-3 expressing a thought, 1 02-3 extensionlintension, 1 48, 152 falsity, 1 8 0-90, 1 98 family resemblance, 329, 33 1-2 Field, H. H., 293n Fleming, P., 337 Fodor, ]., 9- 1 0, 272, 292n, 296n, 321, 326, 347 force, 7, 1 1 - 1 2, 47- 1 20, 359 assertoric, 54-9, 76, 78 imperative, 62 interrogative, 58 formal logic, mathematization of, 30-9 Frege, G., 2-3, 4n, 9, 1 9-20, 29, 32-6, 3 8-44, 50, 59n, 63, 70, 72, 74n, 9 1 n, 1 05, 1 22, 1 24-6, 130- 1 , 1 3 3 , 135, 147n, 1 52-3, 156, 1 63-5, 1 73-4, 1 78-9, 1 86, 200, 208n, 2 1 1 n, 2 13n, 2 1 7, 244-5, 300-5, 3 13, 3 1 7-21, 334, 356, 3 80- 1 his logical system, 33-9 his theory of assertion, 54-9 function/argument analysis, 32-7, 3 1 7 function, mathematical, 32-3, 266 Galilean style, 283-6, 299, 307, 309 Galileo, 1 4 Gazdar, G., SOn generality, 1 79, 246 logic of, 33-4 see also quantification theory gestures, 1 1 0- 1 1 , 1 84, 376 Godel, K., 1 3 1 , 1 40 grammar, traditional, 5 1 , 275 Hacker, P. M. S., 13n, 1 9n, 35n, 64n, 1 42n, 153n, 2 1 7n, 253n, 279n, 320n, 332n, 348n, 349n Hacking, I., 1 6n

Hiigerstrom, A., 257n Hamilton, W., 28 Hamilton, W. R., 30 Hare, R. M., 6 1 -4, 72, 8 1 Harris, R., 7n, 1 7n, 286n, 358 Hart, H. L. A., 4, 1 92n Hilbert, D., 1 30, 148 Hobbes, T., 1 7, 43n homunculus fallacy, 342-3 Hume, D., 1 6n, 343-4 Huntingdon, E. Y., 32 Husser!, E. G. A., 29n

'I

:\1� ,\� 1\1:

I

393

isomorphism of language and reality, 41 Jevons, W . S., 3 1-2 Johnson, S., 68-9 judgment, 27, 29, 33, 55-7 synthetic conception of, 3 1 3 kaleidoscopic mind, 1 8 - 1 9, 135, 3 1 7, 338 Kant, I., 2 7 Katz, ]. ]., 9- 1 0, 50, 66, 68, 8 1 , 1 8 1 , 272, 3 0 1 -2, 3 10n, 3 2 1 , 326-8 , 335, 347 Kenny, A. ]. P., 63n, 65n, 71, 8 1 , 94, 1 0 7n, 1 1 6n, 305n, 342, 350n Kepler, J., 1 4, 308 Kircher; 23 Kneale, W. and M., 336 knowing a language, 267, 276- 8 1 , 344, 3 60 Kripke, S., 157, 22 1n

idea, 1 5 - 1 8 , 22, 27, 3 1 7 complex, 1 8, 27 innate, 290n simple, 1 8, 27, 3 1 7 idealization in linguistics, 283-6 ideal speaker/hearer, 68, 267, 283-6, 3 12, 3 6 1 illocutionary force, 60- 1, 65, 70, 71n, 74, 1 1 6, 155 illocutionary force potential, 65-70 imagination, 339 imperative inference, 62-3 Lakoff, G., 277n, 289n, 299n imperatives, 84, 94-6, 98 Langendoen, T. D., 74, 89n, 271, 284n indefinables, 1 8, 141 -2, 233, 329-30, language: 332 boundaries of, 1 84, 204-5, 376-7 independence proofs, 1 3 1 as a calculus, 1 1 , 46-8, 125, 245, indeterminacy of translation, 159, 304-5, 357 2 1 7- 1 8 capacity, 280 indexicals, 155, 1 97-201, 2 1 8n as a guide to thought, 1, 2 1 indicative core, 88, 97-8, 1 03 ideal, 382 indirect speech, 5 1-2, 28, 77-8, nature of, 7-8 1 06- 12, 1 1 7- 1 8, 1 52, 223 obscures thought, 20, 37 infinity of sentences (of a language), philosophy of, 2, 5, 3 86-9 1 60, 271, 3 0 1 , 306, 32 1-3, 346, as a social practice, 285 353, 358-9, 365-8 as the subject-matter of philosophy, innate knowledge, 290, 3 1 1 , 341 2-3 intension, 150-2, 220 as a system, 1 1, 47, 374-5 intensional isomorphism, 2 1 1 of thought, 1 6 1 n intensional logic, 147-53 language/a language, 375-6 internal relations, 65n, 79, 92-3, langue, la, 7, 67, 267, 270- 1 , 281 99- 1 02, 1 07-8, 1 1 1, 1 16- 1 8, learning a language, 276-7, 287-92, 1 78, 1 8 7, 23 1 , 233, 239-40, 264, 346 297-8, 3 78 Leibniz, G. W., 3, 22-4, 27, 32, 149 interrogatives, 53, 74, 1 00- 1 Lemmon, E. ]., 200n intuitionistic logic, 156, 1 70n Lewis, C. I., 149

394

Index

Lewis, D., 53, 73, 122, 150-2, 157, 159, 1 6 1 lexicon, 7 liar paradox, 1 8 1 linguistic intuitions, 284n, 3 14n, 325 linguistics, 9- 1 1 diachronic, 7, 267-75 subject-matter of, 7-8 synchronic, 7, 267-75 Locke, J., 1 7, 20- 1 , 26, 341 logic: axioms of, 35, 44, 130 laws of, 28-9, 35, 130- 1, 244, 302-4 normative science of, 28, 244 subject matter of, 25-6, 29-33, 39, 140- 1 , 1 65 logical: algebra, 30-3 atomism, 1 35-9, 3 8 1 calculi, 2 , 3 0 connectives, 34, 41-2, 1 15, 126-7, 1 34, 1 70-9 form, 9, 41, 178-9, 246, 3 1 8 necessity, 29, 43, 46 positivism, 59, 6 1-2, 139-43, 156, 1 77, 2 1 9 semantics, 126-32, 140-7 syntax, 41, 43, 127, 140-8, 246 truths, 35, 43-4, 86, 127-9, 140, 147-50 logicism, 38 Lull, R., 23 Lyons, J., 67n, 271

Macbeth, 226 Malcolm, N., 1 1 0

mancipatie, 257n Matthews, P. H., 327 meaning, circumstance-dependence of, 2 1 8-28 mental organ, language as, 248, 280- 1 mental state, 279-80, 282, 288, 295, 347-8 Mercator projection, 85, 357 metalanguage, 143-4, 148-9 metalogic, 130 metaphor, 1 85n, 222, 227-8

f �

:f:

Index

i:,

modal logic, 149-50 model theory, 122, 128-30, 145, 150, 234-5 mood (of verb), 5 1 -4, 71n imperative, 94 indicative, 54, 56 mood-setter, 89n, 97-9 naming, 85 neustic, 63-4, 72-3, 96 Newton, 1., 14, 227, 297 Nielsen, H. A., 290n normative: concepts, 250 explanation, 257-9 phenomena, 256-62 statements, 252, 254 numbers, 1 86, 245, 255n, 273, 304 objects of comparison, 283-4 occasion sentences, 157 Olivecrona, K., 257n onus of match, 71n open texture, 224n optatives, 84 orders, 62-3, 94-6 ostensive definitions, 1 14, 141-6, 329 paintings, understanding of, 355-6 paradigms, 227 paraphrase, 87- 1 06, 1 1 3, 2 1 6- 1 7, 230- 1 parole, la, 7, 67, 267, 271, 281 Peacocke, C., 264n Peirce, c., 32 performance, 8, 67, 70, 28 1-2, 3 1 0- 1 1 performatives, 73 Piaget, J., 291n philosophy: of language, 3 86-9 linguistic, 3 method in, 3 83-6 task of, 2-5, 1 69 phrastic, 63-4, 68, 75, 78, 8 1, 96, 103 picture theory of meaning, 39-46, 64, 3 1 8- 1 9

Plato, 227, 245, 304, 341 Platonism, 20, 28-9, 1 1 7- 1 8, 207, 244-5, 263-5, 297, 300-7 Port Royal Grammar, 54 Port Royal Logic, 17-18, 20-2, 26-8 possibility, 1 5 7 possible word semantics, 125, 147-53, 1 5 7, 2 1 0- 1 1 Postal, P. M., 328 Potts, T., 145 pragmatics, 49-50, 67-8, 154, 171, 1 84, 1 98-9, 201-2 predicate calculus, 33, 128-30, 380- 1 Principia Mathematica, 126, 1 3 1 , 133, 156, 357 Prior, A. N., 262n proof theory, 1 28-9 proposition, 39, 59, 8 1 , 101-2 propositional calculus, 1 3 1 psychologism, 28-9, 32, 1 60-2, 270- 1, 302, 305-6 Putnam, H., 221n

generality of, 251-2 as guides, 25 1, 259-60 hidden, 249, 266, 288, 293, 298-9, 3 13, 334, 342, 344 innate knowledge of, 287 language-dependence of, 253-4, 356 mechanical mythology of, 294-300 necessary, 249 physical realization of, 294, 296, 298-9 Platonic mythology of, 300-7 private, 293-4 psychological mythology of, 286-94 and reasons, 259-60 and regularities, 25 1, 256 semantic field of, 250 source of, 25 1 , 254, 262 as standards, 25 1, 259-60, 265 Russell, B., 2-3, 37-40, 42, 54, 59, 125, 1 35-9, 149, 203n, 306, 3 1 7, 3 8 1-2 Ryle, G., 4, 341

quantifiers, 33, 128, 1 70, 179, 246-7 questions: sentence-, 58, 63, 84 Wh-, 84, 1 00- 1 Quine, W. O., 125, 157, 159, 1 89, 1 95, 2 1 On, 2 1 7, 22 1n

samples, 146, 1 84, 204n, 330 Saussure, F. de, 7-9. 1 8, 67, 267-71, 2 8 1 , 325 Schlick, M., 2-3, 123, 1 39n, 142, 1 5 6n, 2 1 9n, 320- 1 Schroder, E., 32 scientific method in linguistics, 307-15 Searle, J., 65-6, 75, 81, 1 07n, 108n, 1 1 6n self-evidence, 130 semantic: markers, 328 mood, 53, 64, 71, 81, 87 semantics, 6, 67-8, 1 15, 140-7, 152, 153-67 truth-conditional, 36, 48-50, 59-61, 66, 68, 72-3, 87, 108, 1 15, 1 2 1-242 semantics/pragmatics distinction, 50, 54, 1 54 sense, 7, 1 1-2, 47-120, 239n determinacy of, 2 1 7 Frege's concept of, 34, 152, 208n, 2 1 ln, 2 13n identity of, 153

Radford, A., 339 radical translation, 1 96n rationalism, 1 6 realism/anti-realism, 125, 155-6 recognition, 12, 66, 1 09, 1 1 1- 12, 160, 231, 232n, 355, 370 Reid, T., 24 remember, 367-8 representation, forms of, 379-86 Ross, J. R., 73n, 89n rule-following, 254-5, 259-60, 297-8 rule-formulation, 1 87, 252-6, 260, 263, 265, 296 rules, 1 87, 243-3 15, 373 consequences of, 263-5 as definitory, 259, 261, 264 existence of, 262-6, 271-5

I·'.�·. '

.

>: ,

,

395

·f ", . : .

396

Index

senselforce distinction, 7, 1 1 - 12, 36-7, 47- 120 sense/nonsense distinction, 154, 224-8, 336-8 sense/reference distinction, 33-4, 40, 153 sentence: constructional history of, 247, 249 forms of, 5 1-5, 59, 67, 83-4, 89-90 meaning of, 121-4, 1 68, 1 8 9-205, 207-32, 334-9 non-assertoric, 48-50 primacy of, 229-30 use of, 5 1 , 89-91 sentence-radical, 53, 64, 71-6, 78, 81, 1 0 1 -4, 1 13, 190 Skinner, B. F., 290, 345 Smith, N., 3 74 Smith, S., 226, 337 Socrates, 1 speech-acts, 49-52, 57-63, 66, 70- 1, 77-9, 83n, 89-90, 107, 359, 370 speech-act theories of meaning, 64-7, 1 15 - 1 9 speech circuit, 1 8, 268 Sraffa, P., 1 1 0 state of affairs, 39-41, 45, 8 1 , 85-6 Stenius, E., 64, ?ln, 72, 8 1 Stoppard, T., 345 Strawson, P. F., 198n surface structure, 8 syllogism, 26 symbol, 3 77 synchronic! diachronic distinction, 267-75 synonymy, 97-8, 207- 1 8, 23 1 syntax, 6, 67, 154 T-sentence, 123, 143-6, 149, 1 88-9, 1 93 -6, 208- 10, 220, 239-40, 329, 357-8 tacit knowledge, 46, 176, 275, 289, 2 9 1 , 295-6, 305, 3 1 6, 339-45, 361 Tarski, A., 47, 123, 126, 130n, 140-7, 149, 153, 157, 1 64, 1 80-2, 193 T arskian semantics, 140-7 tautology, 140

.\:

i

·

Index

theory of meaning, 4-7, 9-13, 25, 47-50, 1 08-9, 12 1 -242, 276, 322, 336, 345, 357-9, 367-8, 371-9 thought: . correspondence with language, 22-5, 36-7 in Frege, 34, 56 independence of language, 17- 1 8, 20- 1 language of, 292-3 laws of, 29 thought-building-blocks, 320 Toulmin, S., 8 1n trace theory, 248 Tractatus Logico-Philosophicus, 2-5, 8, 39-46, 47, 54, 64, 1 10, 122, 1 26-9, 1 35-8, 140-2, 148, 150, 1 6 1 , 164, 1 70, 1 74, 1 76, 191, 1 96, 2 1 1 n, 2 1 7, 2 1 9, 245-6, 305':"" 6, 3 1 8-21, 356, 3 80-5 truth, 12, 156, 1 80-90, 198 as a property of sentences, 1 82-90, 200 truth conditions, 6-7, 1 1 - 12, 42, 46, 48-9, 72, 12 1 -242 truth definition for a language, 1 80-2 truth-functions, 34 truth-tables, 42, 126-7, 136, 1 70, 1 74-8, 193 type-sentences: as bearers of force, 49, 69-70, 90 as bearers of meaning, 6, 90, 108, 1 87, 1 9 8-20 1 , 220-3, 228

possibility of, 12, 346-56 of sentences, 238-41 theory of, 1 09, 1 60, 237, 276, 340 vehicle of, 352 of words, 238-9 universal grammar, 248, 287-8, 290-2, 3 1 1 utterer's meaning, 21 6n, 222-3 vagueness, 2 1 3n validity, semantic conception of, 63, 1 27-8, 130, 147, 15 1-2 van Buren, P., 292 Venn, ]., 32 Vennemann, T., 277 verifiability, principle of, 59, 122-3 Vienna Circle, 2-3, 122-3, 139, 141, 305,·321 von Wright, G. H., 252n

underdetermination, 158 understanding, 6, 1 0, 4 1 , 45, 65-8, 77, 1 09, 1 1 1 - 12, 1 1 9, 16 1-2, 1 89, 230, 233, 238-41, 246, 267, 295, 3 03-4, 309, 3 1 1, 378 as an ability, 347-52, 354 criteria of, 240- 1, 350, 354 explanatory theory of, 324-56, 340 generative theory of, 3 1 6-68 minimalist explanation of, 322, 324, 326, 340n, 356-68 of new sentences, 1 1, 41, 45, 160, 3 1 6-2 1 , 345, 56

.' ,

I'�

.,

.. .

397

Waismann, F., 63n, 2 1 9, 224n Whateley, R., 25 Whitehead, A. N., 2, 3 17 Wilkins, ]., 23 Wilson, D., 374 Wittgenstein, L., 2-4, 1 9, 37n, 39-46, 54-5, 59, 63-4, 85, 105, 106, 1 1 0, 1 22-3, 125-9, 135-40, 144, 148, 153, 16 1n, 164-5, 1 70, 1 77, 1 79, 1 9 1 , 2 1 6, 2 1 9-20, 224n, 239, 242, 264, 294n, 3 05-6, 3 12, 339, 346, 354n, 363-4, 3 90-5 word-meaning, 328-33 Wordsworth, W., 227 Zeno's paradox, 90, 353n